SigLIP 2 Desktop
Compare a still image with your candidate text labels locally using Google SigLIP2 Base/16 at224pixels. Returns complete ranked label–image matching scores.
Desktop app
Run SigLIP 2 Desktop on your computer
Download the desktop app to load this model, provide input and view results locally.
Public desktop downloads are coming soon.Device requirements
If your device falls short
MemoryAdvisory only: close other apps, reduce image/tile/token/batch settings or choose a smaller verified model/precision. CPU can need more RAM than GPU.
GPUIf the selected GPU runtime fails, existing CPU fallback may recover when enabled, but CPU has its own memory/runtime needs.
PlatformUse a compatible OS/architecture/runtime or another model. These estimates do not add hardware detection or block runs.
Benchmark evidence
Saved desktop measurements apply to the recorded hardware and runtime. Fallback cases report the device that actually ran; they do not prove GPU support on the requested device.
SigLIP 2 Base — Image & Text Matching (224) Desktop
145 configurations testedQuality criteria
{ "minimumTop1Accuracy": 1, "maximumCpuScoreDifference": 0.03, "finiteScoresInUnitInterval": true, "scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification", "performanceTargets": { "goodInferenceMs": 250, "acceptableInferenceMs": 1000, "description": "One224x224 image and four labels; good below250ms, acceptable below1s, slower cases retained but not recommended" }, "version": 2, "minimumLogOddsMargin": 1, "gateRevisionReason": "V1 required0.05absolute sigmoid margin, which conflated calibrated confidence with ranking despite the native independent sigmoid contract. All six semantic fixtures ranked correctly but dog/portrait failed absolute margin. Preserve fullv1failures;v2requires expected top1plus ≥1natural-log-odds separation (odds ratio≥e), finite scores and unchanged0.03CPU agreement. This measures ranking separation without treating sigmoid scores as calibrated probabilities." }
| Precision | Device | Result | Platform | Inference / load | |
|---|---|---|---|---|---|
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 17 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 17 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 17 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 17 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 19 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 19 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 19 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 19 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 19 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 19 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 20 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 22 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 24 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 25 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 93 ms / 903 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 95 ms / 933 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 96 ms / 887 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 96 ms / 838 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 100 ms / 907 ms |
BF16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 100 ms / 859 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 107 ms / 900 ms |
BF16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 113 ms / 905 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 125 ms / 1107 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 82 ms / 0 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 82 ms / 0 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 82 ms / 0 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 83 ms / 0 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 83 ms / 0 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 83 ms / 0 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 83 ms / 0 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 84 ms / 0 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 84 ms / 0 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 84 ms / 0 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 84 ms / 0 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 84 ms / 0 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 85 ms / 0 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 85 ms / 0 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 85 ms / 0 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 85 ms / 0 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 85 ms / 0 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 85 ms / 0 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 85 ms / 0 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 85 ms / 0 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 87 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 136 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 136 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 136 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 136 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 137 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 137 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 139 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 139 ms / 632 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 139 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 139 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 139 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 140 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 140 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 140 ms / 656 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 141 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 141 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 141 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 141 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 142 ms / 733 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 142 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 142 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 142 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 143 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 143 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 191 ms / 623 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 195 ms / 694 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 195 ms / 709 ms |
BF16 | mps | Failed | — | macos · aarch64 | 18 ms / 0 ms |
Test setup & sample details
- Quality gate
- { "minimumTop1Accuracy": 1, "maximumCpuScoreDifference": 0.03, "finiteScoresInUnitInterval": true, "scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification", "performanceTargets": { "goodInferenceMs": 250, "acceptableInferenceMs": 1000, "description": "One224x224 image and four labels; good below250ms, acceptable below1s, slower cases retained but not recommended" }, "version": 2, "minimumLogOddsMargin": 1, "gateRevisionReason": "V1 required0.05absolute sigmoid margin, which conflated calibrated confidence with ranking despite the native independent sigmoid contract. All six semantic fixtures ranked correctly but dog/portrait failed absolute margin. Preserve fullv1failures;v2requires expected top1plus ≥1natural-log-odds separation (odds ratio≥e), finite scores and unchanged0.03CPU agreement. This measures ranking separation without treating sigmoid scores as calibrated probabilities." }
- Tested
- 2026-11-02T15:56:55.653243+00:00
- Environment
- { "hardware": "Apple M3 Max", "os": "macOS-27.2-arm64-arm-64bit-Mach-O", "runtime": "Transformers5.22.1/Torch2.10.0/Pillow12.0.0", "interpreterScope": "Existing isolated private probe venv from app-ownedCPython3.13.18; managed packaged provision is separate evidence" }
- Runtime
- transformers 5.22.1
- Artifact revision
- 75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2
- Fixture
{ "file": "model-samples/desktop/dinov2/dinov2-dog.png", "size": 709899, "sha1": "409baef01589c0a55c12bba3b4f5330adc74b1c7", "sha256": "cc986ccf120f7892c9ee0202001c30a91a3039cc5859d851b981476507a25fe3", "url": "https://assets.localai.tools/samples/v1/files/cc986ccf120f7892c9ee0202001c30a91a3039cc5859d851b981476507a25fe3/dinov2-dog.png" }- Rerun when
- Original weights/processor/tokenizer artifact revision; Pinned runtime or relevant OS/driver/backend behavior; Fixture bytes/labels/expected result or quality gate; Image preprocessing/prompt template/sigmoid scoring
Per-case quality & failure details
[
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 693.5665829951176,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 213.1108750036219,
"inferenceMs": 194.72662499174476,
"imagesPerSecond": 5.135404570599393,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2270.0171249889536,
"totalMs": 906.6774579987396,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453375488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.013536862301407382,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.701620156859222,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 153.45925001020078,
"inferenceMs": 142.83087498915847,
"imagesPerSecond": 7.0012873622450655,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 153.45925001020078,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453375488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0004604809237207519,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 6.372386985573321,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 155.34804100752808,
"inferenceMs": 142.41220799158327,
"imagesPerSecond": 7.021869923251953,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 155.34804100752808,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453375488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.000999590574224385,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 6.417356493634003,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 153.01145799458027,
"inferenceMs": 140.76875000318978,
"imagesPerSecond": 7.103849398231783,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 153.01145799458027,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453375488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.48676048092602286,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 8.83504673136097,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 148.77445800811984,
"inferenceMs": 140.95070799521636,
"imagesPerSecond": 7.094678801002819,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 148.77445800811984,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453375488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.4216608988135704,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 8.74613371758602,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 147.2490409942111,
"inferenceMs": 139.8749579966534,
"imagesPerSecond": 7.149242540068721,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 147.2490409942111,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453375488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0007559390214737505,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 1.364363714701442,
"passed": true,
"top1Accuracy": null
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 141.47991700156126,
"inferenceMs": 139.54637499409728,
"imagesPerSecond": 7.166076510710503,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 141.47991700156126,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453375488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 153.19808399362955,
"inferenceMs": 142.67633299459703,
"imagesPerSecond": 7.0088709109020115,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 153.19808399362955,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453375488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 152.48366601008456,
"inferenceMs": 141.94966699869838,
"imagesPerSecond": 7.0447505876091245,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 152.48366601008456,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453375488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 150.98820799903478,
"inferenceMs": 140.54799999576062,
"imagesPerSecond": 7.115006972921445,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 150.98820799903478,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453375488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 152.16733299894258,
"inferenceMs": 141.01812501030508,
"imagesPerSecond": 7.0912870237561565,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 152.16733299894258,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453375488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 152.44116699614096,
"inferenceMs": 141.74829100375064,
"imagesPerSecond": 7.0547587764112105,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 152.44116699614096,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453375488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 623.3401249919552,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 208.02629200625233,
"inferenceMs": 191.4142500027083,
"imagesPerSecond": 5.224271442621703,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2176.693832996534,
"totalMs": 831.3664169982076,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1437712384
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 147.3222920030821,
"inferenceMs": 136.91308301349636,
"imagesPerSecond": 7.303903892818073,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 147.3222920030821,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1437712384
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 146.97449999221135,
"inferenceMs": 136.5014580078423,
"imagesPerSecond": 7.325929075003344,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 146.97449999221135,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1437712384
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 146.24345800257288,
"inferenceMs": 135.98420799826272,
"imagesPerSecond": 7.353795081946395,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 146.24345800257288,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1437712384
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 146.0283329943195,
"inferenceMs": 135.96945800236426,
"imagesPerSecond": 7.354592823210429,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 146.0283329943195,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1437712384
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 145.89658299519215,
"inferenceMs": 135.787999999593,
"imagesPerSecond": 7.364421009242328,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 145.89658299519215,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1437712384
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 708.8267080107471,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 214.08420799707528,
"inferenceMs": 195.00583301123697,
"imagesPerSecond": 5.128051733418539,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2372.8068329946836,
"totalMs": 922.9109160078224,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1440055296
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 149.89566699659918,
"inferenceMs": 139.13412499823608,
"imagesPerSecond": 7.187309367940308,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 149.89566699659918,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1440055296
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 150.2373749972321,
"inferenceMs": 139.36549999925774,
"imagesPerSecond": 7.175376976406112,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 150.2373749972321,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1440055296
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 149.3296660046326,
"inferenceMs": 138.53412499884143,
"imagesPerSecond": 7.218438056387645,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 149.3296660046326,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1440055296
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 147.1184169931803,
"inferenceMs": 136.46074999996927,
"imagesPerSecond": 7.328114494462512,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 147.1184169931803,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1440055296
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 149.49887500551995,
"inferenceMs": 138.8800840068143,
"imagesPerSecond": 7.200456474024987,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 149.49887500551995,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1440055296
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 733.4511250082869,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 159.1081250080606,
"inferenceMs": 141.6552090086043,
"imagesPerSecond": 7.059394476197899,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2382.35391699709,
"totalMs": 892.5592500163475,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1451261952
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.013536862301407382,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.701620156859222,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 98.16754100029357,
"inferenceMs": 87.22954199765809,
"imagesPerSecond": 11.464006082100576,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 98.16754100029357,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1451261952
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0004604809237207519,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 6.372386985573321,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 97.46550000272691,
"inferenceMs": 85.09654100635089,
"imagesPerSecond": 11.751358964465647,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 97.46550000272691,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1451261952
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.000999590574224385,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 6.417356493634003,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 96.95733299304266,
"inferenceMs": 84.91208399937022,
"imagesPerSecond": 11.776886785718471,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 96.95733299304266,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1451261952
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.48676048092602286,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 8.83504673136097,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 92.22350000345614,
"inferenceMs": 84.73766699898988,
"imagesPerSecond": 11.801127354756186,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 92.22350000345614,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1451261952
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.4216608988135704,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 8.74613371758602,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 92.5217499898281,
"inferenceMs": 85.40695800911635,
"imagesPerSecond": 11.708647905399696,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 92.5217499898281,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1451261952
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0007559390214737505,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 1.364363714701442,
"passed": true,
"top1Accuracy": null
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 87.03591699304525,
"inferenceMs": 85.05066600628197,
"imagesPerSecond": 11.757697463840419,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 87.03591699304525,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1451261952
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 96.24908299883828,
"inferenceMs": 85.43150000332389,
"imagesPerSecond": 11.70528435016467,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 96.24908299883828,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1451261952
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 95.34529100346845,
"inferenceMs": 84.46491700306069,
"imagesPerSecond": 11.839234980409248,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 95.34529100346845,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1451261952
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 95.96512500138488,
"inferenceMs": 84.91441700607538,
"imagesPerSecond": 11.776563218098204,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 95.96512500138488,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1451261952
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 95.01445801288355,
"inferenceMs": 84.38275000662543,
"imagesPerSecond": 11.850763336362984,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 95.01445801288355,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1451261952
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 95.9624170063762,
"inferenceMs": 84.89025000017136,
"imagesPerSecond": 11.779915832477597,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 95.9624170063762,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1451261952
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 632.0584999921266,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 156.23258300183807,
"inferenceMs": 138.84100000723265,
"imagesPerSecond": 7.202483415906735,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2147.9964580066735,
"totalMs": 788.2910829939647,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1434124288
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 94.15795799577609,
"inferenceMs": 83.65137499640696,
"imagesPerSecond": 11.95437612403804,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 94.15795799577609,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1434124288
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 94.50454100442585,
"inferenceMs": 83.27737500076182,
"imagesPerSecond": 12.008063414473042,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 94.50454100442585,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1434124288
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 94.27600000344682,
"inferenceMs": 83.79387500463054,
"imagesPerSecond": 11.934046491402134,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 94.27600000344682,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1434124288
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 92.98679100174922,
"inferenceMs": 82.73587499570567,
"imagesPerSecond": 12.086655275621416,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 92.98679100174922,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1434124288
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 93.55166699970141,
"inferenceMs": 83.15158299228642,
"imagesPerSecond": 12.026229255223743,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 93.55166699970141,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1434124288
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 655.9444999875268,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 158.83791699889116,
"inferenceMs": 140.44425000611227,
"imagesPerSecond": 7.120263022206171,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2142.452582993428,
"totalMs": 814.782416986418,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1438187520
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 92.54024999972899,
"inferenceMs": 82.01529200596269,
"imagesPerSecond": 12.192848132849393,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 92.54024999972899,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1438187520
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 93.42599999217782,
"inferenceMs": 82.42395900015254,
"imagesPerSecond": 12.132394659641978,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 93.42599999217782,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1438187520
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 93.11862500908319,
"inferenceMs": 82.83029099402484,
"imagesPerSecond": 12.072878025650512,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 93.11862500908319,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1438187520
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 94.43233300407883,
"inferenceMs": 83.98179199139122,
"imagesPerSecond": 11.907342964323835,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 94.43233300407883,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1438187520
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014865775454381946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666258783021,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 92.3295840038918,
"inferenceMs": 81.81862499623094,
"imagesPerSecond": 12.222156019440147,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 92.3295840038918,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1438187520
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 907.3916250054026,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 128.60254201223142,
"inferenceMs": 99.90416700020432,
"imagesPerSecond": 10.009592492753129,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568489472,
"processInitMs": 2371.7168330040295,
"totalMs": 1035.994167017634,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 2.60770320892334e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.013536888465750962,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.701624001435487,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.36424999765586,
"inferenceMs": 22.744999994756654,
"imagesPerSecond": 43.965706758871285,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568489472,
"processInitMs": 0,
"totalMs": 35.36424999765586,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.326000973582268e-9,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.00046047559845874275,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 6.372376371252769,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 37.33579099935014,
"inferenceMs": 23.193832996184938,
"imagesPerSecond": 43.11490904347229,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568489472,
"processInitMs": 0,
"totalMs": 37.33579099935014,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 7.683411240577698e-9,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0009995828688715847,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 6.417335405251958,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 36.81150000193156,
"inferenceMs": 23.111417001928203,
"imagesPerSecond": 43.26865807996841,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568489472,
"processInitMs": 0,
"totalMs": 36.81150000193156,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000014007091522216797,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.48676188112585805,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 8.835048649255077,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 32.42854199197609,
"inferenceMs": 23.05674999661278,
"imagesPerSecond": 43.371247038151864,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568489472,
"processInitMs": 0,
"totalMs": 32.42854199197609,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000001430511474609375,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.4216623295505997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 8.746141527492373,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 31.79154099780135,
"inferenceMs": 22.63341599609703,
"imagesPerSecond": 44.18246013648328,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568489472,
"processInitMs": 0,
"totalMs": 31.79154099780135,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.1641532182693481e-8,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0007559528748970479,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 1.3643837057765182,
"passed": true,
"top1Accuracy": null
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 26.114167005289346,
"inferenceMs": 22.657624998828396,
"imagesPerSecond": 44.13525248351092,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568489472,
"processInitMs": 0,
"totalMs": 26.114167005289346,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.52287499769591,
"inferenceMs": 23.165292004705407,
"imagesPerSecond": 43.16802912723383,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.52287499769591,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.15487500408199,
"inferenceMs": 22.601709002628922,
"imagesPerSecond": 44.24444186427163,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.15487500408199,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 36.13529200083576,
"inferenceMs": 23.153458008891903,
"imagesPerSecond": 43.19009279805884,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 36.13529200083576,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.13466700678691,
"inferenceMs": 23.48258400161285,
"imagesPerSecond": 42.58475131745796,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.13466700678691,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.138833001838066,
"inferenceMs": 22.95708400197327,
"imagesPerSecond": 43.55953917814846,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.138833001838066,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.62437499931548,
"inferenceMs": 22.902457989403047,
"imagesPerSecond": 43.663435621744156,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.62437499931548,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 37.70845799590461,
"inferenceMs": 24.449416989227757,
"imagesPerSecond": 40.90077078077539,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 37.70845799590461,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.57649999856949,
"inferenceMs": 22.89666600700002,
"imagesPerSecond": 43.6744808040733,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.57649999856949,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.122792003676295,
"inferenceMs": 22.91995800624136,
"imagesPerSecond": 43.630097390566284,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.122792003676295,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 36.064666986931115,
"inferenceMs": 23.499832997913472,
"imagesPerSecond": 42.55349389456466,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 36.064666986931115,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.555167007260025,
"inferenceMs": 22.98158399935346,
"imagesPerSecond": 43.51310162206978,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.555167007260025,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.41920900170226,
"inferenceMs": 22.98875000269618,
"imagesPerSecond": 43.49953781230895,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.41920900170226,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.53254099097103,
"inferenceMs": 22.980084002483636,
"imagesPerSecond": 43.51594188654499,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.53254099097103,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.37337500893045,
"inferenceMs": 22.91062500444241,
"imagesPerSecond": 43.64787079383903,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.37337500893045,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 34.730207989923656,
"inferenceMs": 22.589583008084446,
"imagesPerSecond": 44.26819209730946,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 34.730207989923656,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.295542009407654,
"inferenceMs": 22.880875010741875,
"imagesPerSecond": 43.70462228959909,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.295542009407654,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.099083004752174,
"inferenceMs": 22.536417003721,
"imagesPerSecond": 44.37262586305931,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.099083004752174,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 36.2060829938855,
"inferenceMs": 22.59599999524653,
"imagesPerSecond": 44.25562047310886,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 36.2060829938855,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 34.87620899977628,
"inferenceMs": 22.28849999664817,
"imagesPerSecond": 44.86618660521721,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 34.87620899977628,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.79820900631603,
"inferenceMs": 22.717375002685003,
"imagesPerSecond": 44.019170343484156,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.79820900631603,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.476833989378065,
"inferenceMs": 22.794124990468845,
"imagesPerSecond": 43.87095360835921,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.476833989378065,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.91841700836085,
"inferenceMs": 23.07058300357312,
"imagesPerSecond": 43.34524185388476,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.91841700836085,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.41408399178181,
"inferenceMs": 22.61929100495763,
"imagesPerSecond": 44.21005060595502,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.41408399178181,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.63458300777711,
"inferenceMs": 22.96816599846352,
"imagesPerSecond": 43.53852197284259,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.63458300777711,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.76191699539777,
"inferenceMs": 22.955874999752268,
"imagesPerSecond": 43.56183330022452,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.76191699539777,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.83995799999684,
"inferenceMs": 23.278958004084416,
"imagesPerSecond": 42.95724919579926,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.83995799999684,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.46558400557842,
"inferenceMs": 22.698500004480593,
"imagesPerSecond": 44.05577460196066,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.46558400557842,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.35645800002385,
"inferenceMs": 22.85416700760834,
"imagesPerSecond": 43.75569670367298,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.35645800002385,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.82516600727104,
"inferenceMs": 22.95620799122844,
"imagesPerSecond": 43.56120141366988,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.82516600727104,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.74937500525266,
"inferenceMs": 22.964750009123236,
"imagesPerSecond": 43.544998295332135,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.74937500525266,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 939655168
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 1107.3093330051051,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 159.75891600828618,
"inferenceMs": 125.14933300553821,
"imagesPerSecond": 7.990454091798853,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568489472,
"processInitMs": 3025.4182089993265,
"totalMs": 1267.0682490133913,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 780599296
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 36.550665987306274,
"inferenceMs": 22.91266700194683,
"imagesPerSecond": 43.64398085631117,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568489472,
"totalMs": 36.550665987306274,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 780599296
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 36.74958398914896,
"inferenceMs": 23.08779099257663,
"imagesPerSecond": 43.312935409088205,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568489472,
"totalMs": 36.74958398914896,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 780599296
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.69649999553803,
"inferenceMs": 22.687667005811818,
"imagesPerSecond": 44.07681053075372,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568489472,
"totalMs": 35.69649999553803,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 780599296
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.86341699701734,
"inferenceMs": 22.652957995887846,
"imagesPerSecond": 44.14434530720131,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568489472,
"totalMs": 35.86341699701734,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 780599296
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.895749999326654,
"inferenceMs": 22.6634999999078,
"imagesPerSecond": 44.12381141500952,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568473088,
"totalMs": 35.895749999326654,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 780599296
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 933.2280420057941,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 123.16941599419806,
"inferenceMs": 95.34145900397561,
"imagesPerSecond": 10.488616499547183,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568489472,
"processInitMs": 2591.63824999996,
"totalMs": 1056.3974579999922,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 780894208
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 34.97725000488572,
"inferenceMs": 22.755917001632042,
"imagesPerSecond": 43.94461448986127,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568489472,
"totalMs": 34.97725000488572,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 780894208
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.227250002208166,
"inferenceMs": 22.63829199364409,
"imagesPerSecond": 44.17294380162423,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568489472,
"totalMs": 35.227250002208166,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 780894208
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.263875004602596,
"inferenceMs": 22.810082999058068,
"imagesPerSecond": 43.84026134588351,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568489472,
"totalMs": 35.263875004602596,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 780894208
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.015500005101785,
"inferenceMs": 22.656874993117526,
"imagesPerSecond": 44.13671348338064,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568489472,
"totalMs": 35.015500005101785,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 780894208
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.210719347000122e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01486578751791967,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.815666115290942,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 40.00033299962524,
"inferenceMs": 25.01645899610594,
"imagesPerSecond": 39.973682932331066,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1568489472,
"totalMs": 40.00033299962524,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 780894208
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002952190116047859,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014570974846719764,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.804687483692643,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 899.5774159993744,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 138.56458399095573,
"inferenceMs": 107.23112500272691,
"imagesPerSecond": 9.32565055131679,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"processInitMs": 2411.9644169986714,
"totalMs": 1038.14199999033,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1249558528
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00025771185755729675,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.013279749462526524,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.695312481777884,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 33.46083300129976,
"inferenceMs": 19.63491599599365,
"imagesPerSecond": 50.92968058554679,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"processInitMs": 0,
"totalMs": 33.46083300129976,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1249558528
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000065212370827794075,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.000453960657978314,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 6.359375018789969,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 33.088750002207235,
"inferenceMs": 18.140249987482093,
"imagesPerSecond": 55.126031928449855,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"processInitMs": 0,
"totalMs": 33.088750002207235,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1249558528
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 7.228227332234383e-7,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.000998850669134299,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 6.406249987198913,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 32.09900000365451,
"inferenceMs": 17.380834004143253,
"imagesPerSecond": 57.534638427685316,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"processInitMs": 0,
"totalMs": 32.09900000365451,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1249558528
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.010318934917449951,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.47644198170746677,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 8.796874982972822,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 27.568000004976057,
"inferenceMs": 18.127375005860813,
"imagesPerSecond": 55.16518523375211,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"processInitMs": 0,
"totalMs": 27.568000004976057,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1249558528
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0068780481815338135,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.41478295304114,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 8.718750185738427,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 28.647416998865083,
"inferenceMs": 17.8077089949511,
"imagesPerSecond": 56.15545493715802,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"processInitMs": 0,
"totalMs": 28.647416998865083,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1249558528
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000007343944162130356,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0007532021554652601,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 1.3750000604168866,
"passed": true,
"top1Accuracy": null
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 21.11016599519644,
"inferenceMs": 17.589833005331457,
"imagesPerSecond": 56.851022957233376,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"processInitMs": 0,
"totalMs": 21.11016599519644,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1249558528
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002952190116047859,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014570974846719764,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.804687483692643,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 29.73987499717623,
"inferenceMs": 17.690458000288345,
"imagesPerSecond": 56.52764897232737,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 29.73987499717623,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1249558528
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002952190116047859,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014570974846719764,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.804687483692643,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.72704200167209,
"inferenceMs": 17.787875010981224,
"imagesPerSecond": 56.21806985840955,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 30.72704200167209,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1249558528
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002952190116047859,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014570974846719764,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.804687483692643,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.20637499867007,
"inferenceMs": 17.578540995600633,
"imagesPerSecond": 56.88754261518457,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107099648,
"totalMs": 30.20637499867007,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1249558528
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002952190116047859,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014570974846719764,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.804687483692643,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 31.4608749904437,
"inferenceMs": 17.803333001211286,
"imagesPerSecond": 56.16925774134332,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107099648,
"totalMs": 31.4608749904437,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1249558528
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002952190116047859,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014570974846719764,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.804687483692643,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.123374992399476,
"inferenceMs": 17.902334002428688,
"imagesPerSecond": 55.858638312989626,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107099648,
"totalMs": 30.123374992399476,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1249558528
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002952190116047859,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014570974846719764,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.804687483692643,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 837.7550000004703,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 126.05166700086556,
"inferenceMs": 96.43820799828973,
"imagesPerSecond": 10.36933411306994,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"processInitMs": 2378.2821250060806,
"totalMs": 963.8066670013359,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1231601664
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002952190116047859,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014570974846719764,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.804687483692643,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 31.852917003561743,
"inferenceMs": 18.532250003772788,
"imagesPerSecond": 53.95998865741723,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 31.852917003561743,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1231601664
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002952190116047859,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014570974846719764,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.804687483692643,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.746542004635558,
"inferenceMs": 18.362041999353096,
"imagesPerSecond": 54.460173875826584,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 30.746542004635558,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1231601664
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002952190116047859,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014570974846719764,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.804687483692643,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 29.50391599733848,
"inferenceMs": 17.487499993876554,
"imagesPerSecond": 57.18370266476977,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 29.50391599733848,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1231601664
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002952190116047859,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014570974846719764,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.804687483692643,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.47883399995044,
"inferenceMs": 17.962500001885928,
"imagesPerSecond": 55.671537920390115,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 30.47883399995044,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1231601664
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002952190116047859,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014570974846719764,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.804687483692643,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.77787499933038,
"inferenceMs": 18.149416995584033,
"imagesPerSecond": 55.09818856679043,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 30.77787499933038,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1231601664
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002952190116047859,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014570974846719764,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.804687483692643,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 887.3835830017924,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 124.68487500154879,
"inferenceMs": 95.88766700471751,
"imagesPerSecond": 10.428869856128648,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"processInitMs": 2573.4208329959074,
"totalMs": 1012.0684580033412,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1218412544
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002952190116047859,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014570974846719764,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.804687483692643,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.481207999400795,
"inferenceMs": 17.907375004142523,
"imagesPerSecond": 55.84291387032825,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 30.481207999400795,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1218412544
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002952190116047859,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014570974846719764,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.804687483692643,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.740917005459778,
"inferenceMs": 18.1646250130143,
"imagesPerSecond": 55.05205856347356,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 30.740917005459778,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1218412544
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002952190116047859,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014570974846719764,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.804687483692643,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.15900000173133,
"inferenceMs": 17.456499990657903,
"imagesPerSecond": 57.28525194255231,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 30.15900000173133,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1218412544
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002952190116047859,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014570974846719764,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.804687483692643,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.307332999655046,
"inferenceMs": 17.941333004273474,
"imagesPerSecond": 55.737218620367194,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 30.307332999655046,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1218412544
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002952190116047859,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014570974846719764,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.804687483692643,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.237332990509458,
"inferenceMs": 17.80208399577532,
"imagesPerSecond": 56.17319861187678,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 30.237332990509458,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1218412544
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0020040255039930344,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.016863471220858628,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.812500083562882,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 905.0706670095678,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 144.68866601237096,
"inferenceMs": 112.54133300099056,
"imagesPerSecond": 8.885624270961836,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"processInitMs": 2480.6126250041416,
"totalMs": 1049.7593330219388,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1346158592
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0013742269948124886,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014905643973179394,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.687500045866167,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 37.688249998609535,
"inferenceMs": 22.548332999576814,
"imagesPerSecond": 44.349176500931044,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"processInitMs": 0,
"totalMs": 37.688249998609535,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1346158592
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00002658794983290136,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.00048702559189450767,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 6.375000046260032,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 34.653790993615985,
"inferenceMs": 18.795208001392893,
"imagesPerSecond": 53.20505098564969,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"processInitMs": 0,
"totalMs": 34.653790993615985,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1346158592
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00009750388562679291,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0010969707238928095,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 6.437500036969046,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 32.973415989545174,
"inferenceMs": 17.77758399839513,
"imagesPerSecond": 56.25061313676115,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"processInitMs": 0,
"totalMs": 32.973415989545174,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1346158592
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.013101428747177124,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.49986864819948096,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 8.93749997370247,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 28.440584006602876,
"inferenceMs": 18.27370800310746,
"imagesPerSecond": 54.72343105350864,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"processInitMs": 0,
"totalMs": 28.440584006602876,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1346158592
}
},
{
"precision": "bf16",
"device": "mps",
"status": "fail",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.04423624277114868,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.37743178087839624,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 8.62500006165855,
"passed": false,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 27.540041992324404,
"inferenceMs": 17.78520899824798,
"imagesPerSecond": 56.22649697838861,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"processInitMs": 0,
"totalMs": 27.540041992324404,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1346158592
},
"failure": "Frozen classification/numerical-agreement quality gate failed; unsupported"
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00030601717298850417,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.00047916360199451447,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 1.1249999503263908,
"passed": true,
"top1Accuracy": null
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 21.19016699725762,
"inferenceMs": 17.30816600320395,
"imagesPerSecond": 57.77619649678007,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"processInitMs": 0,
"totalMs": 21.19016699725762,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1346158592
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0020040255039930344,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.016863471220858628,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.812500083562882,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 31.266250007320195,
"inferenceMs": 18.488542002160102,
"imagesPerSecond": 54.087553246933446,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 31.266250007320195,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1346158592
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0020040255039930344,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.016863471220858628,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.812500083562882,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.83712499937974,
"inferenceMs": 18.242416001157835,
"imagesPerSecond": 54.81730051197882,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 30.83712499937974,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1346158592
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0020040255039930344,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.016863471220858628,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.812500083562882,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.264458997407928,
"inferenceMs": 17.84116700582672,
"imagesPerSecond": 56.05014513195306,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107099648,
"totalMs": 30.264458997407928,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1346158592
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0020040255039930344,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.016863471220858628,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.812500083562882,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 29.870041995309293,
"inferenceMs": 18.160790990805253,
"imagesPerSecond": 55.06368089948817,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107099648,
"totalMs": 29.870041995309293,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1346158592
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0020040255039930344,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.016863471220858628,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.812500083562882,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.677749993628822,
"inferenceMs": 17.81570800812915,
"imagesPerSecond": 56.13024189348573,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107099648,
"totalMs": 30.677749993628822,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1346158592
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0020040255039930344,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.016863471220858628,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.812500083562882,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 902.6983329968061,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 120.25920899759512,
"inferenceMs": 92.88416599156335,
"imagesPerSecond": 10.766097637037833,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"processInitMs": 2526.4341669972055,
"totalMs": 1022.9575419944013,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1236959232
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0020040255039930344,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.016863471220858628,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.812500083562882,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 31.549417006317526,
"inferenceMs": 18.525708001106977,
"imagesPerSecond": 53.97904360471656,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 31.549417006317526,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1236959232
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0020040255039930344,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.016863471220858628,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.812500083562882,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.483167007332668,
"inferenceMs": 17.519250002806075,
"imagesPerSecond": 57.080069057741,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 30.483167007332668,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1236959232
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0020040255039930344,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.016863471220858628,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.812500083562882,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 31.176834003417753,
"inferenceMs": 17.918124998686835,
"imagesPerSecond": 55.80941086599669,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 31.176834003417753,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1236959232
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0020040255039930344,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.016863471220858628,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.812500083562882,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 32.82883399515413,
"inferenceMs": 19.417499992414378,
"imagesPerSecond": 51.499935645199386,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 32.82883399515413,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1236959232
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0020040255039930344,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.016863471220858628,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.812500083562882,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 35.01620900351554,
"inferenceMs": 19.352000002982095,
"imagesPerSecond": 51.674245548052006,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 35.01620900351554,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1236959232
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0020040255039930344,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.016863471220858628,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.812500083562882,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 859.3341669911752,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 129.92045799910557,
"inferenceMs": 100.21129199594725,
"imagesPerSecond": 9.97891535058187,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"processInitMs": 2363.6167090007802,
"totalMs": 989.2546249902807,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1176027136
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0020040255039930344,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.016863471220858628,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.812500083562882,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 31.245957987266593,
"inferenceMs": 18.40937499946449,
"imagesPerSecond": 54.320149381990916,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 31.245957987266593,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1176027136
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0020040255039930344,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.016863471220858628,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.812500083562882,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 31.322625000029802,
"inferenceMs": 18.316832996788435,
"imagesPerSecond": 54.59459067925848,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 31.322625000029802,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1176027136
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0020040255039930344,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.016863471220858628,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.812500083562882,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 31.302666990086436,
"inferenceMs": 18.585083991638385,
"imagesPerSecond": 53.80659029843018,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 31.302666990086436,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1176027136
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0020040255039930344,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.016863471220858628,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.812500083562882,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.89733399974648,
"inferenceMs": 18.1834160030121,
"imagesPerSecond": 54.99516701561187,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 30.89733399974648,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1176027136
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0020040255039930344,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.016863471220858628,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"logOddsMargin": 5.812500083562882,
"passed": true,
"top1Accuracy": 1
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.757708998862654,
"inferenceMs": 18.28504200966563,
"imagesPerSecond": 54.68951066513226,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
224,
224
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1107116032,
"totalMs": 30.757708998862654,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "75de2d55ec2d0b4efc50b3e9ad70dba96a7b2fa2",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1176027136
}
}
]SigLIP 2 Base · Patch16 · 256 — Image & Text Matching Desktop
153 configurations testedQuality criteria
{ "minimumTop1Accuracy": 1, "maximumCpuScoreDifference": 0.03, "finiteScoresInUnitInterval": true, "scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification", "performanceTargets": { "goodInferenceMs": 250, "acceptableInferenceMs": 1000, "description": "One image at pinned256px FixRes or256patch NaFlex default and four labels; good<=250ms,acceptable<=1000ms" }, "version": 2, "minimumLogOddsMargin": 1, "gateRevisionReason": "V1 required0.05absolute sigmoid margin, which conflated calibrated confidence with ranking despite the native independent sigmoid contract. All six semantic fixtures ranked correctly but dog/portrait failed absolute margin. Preserve fullv1failures;v2requires expected top1plus ≥1natural-log-odds separation (odds ratio≥e), finite scores and unchanged0.03CPU agreement. This measures ranking separation without treating sigmoid scores as calibrated probabilities." }
| Precision | Device | Result | Platform | Inference / load | |
|---|---|---|---|---|---|
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 25 ms / 2806 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 25 ms / 2806 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 25 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 26 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 26 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 26 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 26 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 26 ms / 2806 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 27 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 2806 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 2806 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 37 ms / 2806 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 37 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 39 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 39 ms / 2806 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 2806 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 2806 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 44 ms / 2806 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 44 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 44 ms / 2806 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 44 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 44 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 44 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 45 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 45 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 45 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 46 ms / 2806 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 46 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 47 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 47 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 49 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 49 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 51 ms / 2806 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 59 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 95 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 99 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 100 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 102 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 102 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 104 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 105 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 106 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 109 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 109 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 111 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 111 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 112 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 113 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 116 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 131 ms / 2696 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 131 ms / 2806 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 131 ms / 2696 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 89 ms / 2430 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 89 ms / 2430 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 89 ms / 2430 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 90 ms / 2430 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 90 ms / 2430 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 90 ms / 2430 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 90 ms / 2430 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 90 ms / 2430 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 90 ms / 2430 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 91 ms / 2430 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 91 ms / 2430 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 91 ms / 2430 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 94 ms / 2430 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 96 ms / 4335 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 96 ms / 4335 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 98 ms / 4335 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 99 ms / 4335 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 101 ms / 4335 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 106 ms / 4335 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 122 ms / 4335 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 153 ms / 2430 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 155 ms / 2432 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 156 ms / 3763 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 157 ms / 2432 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 157 ms / 2547 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 157 ms / 3763 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 157 ms / 2547 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 157 ms / 2432 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 158 ms / 2547 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 159 ms / 2547 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 159 ms / 2432 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 159 ms / 2547 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 159 ms / 2432 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 159 ms / 2432 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 159 ms / 2547 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 160 ms / 2432 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 160 ms / 2432 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 160 ms / 2547 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 160 ms / 2432 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 160 ms / 2432 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 160 ms / 2432 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 161 ms / 2547 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 162 ms / 2547 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 162 ms / 2547 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 163 ms / 2432 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 164 ms / 2432 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 166 ms / 2547 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 167 ms / 3763 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 169 ms / 3763 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 172 ms / 3763 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 173 ms / 2547 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 176 ms / 2547 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 182 ms / 4335 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 196 ms / 3763 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 204 ms / 4093 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 207 ms / 4335 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 210 ms / 4093 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 214 ms / 3763 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 215 ms / 4093 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 219 ms / 2432 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 219 ms / 4335 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 227 ms / 2547 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 231 ms / 4093 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 232 ms / 4093 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 233 ms / 4335 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 249 ms / 4093 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 259 ms / 4335 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 261 ms / 3763 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 263 ms / 4335 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 276 ms / 4093 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 280 ms / 3763 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 280 ms / 3763 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 287 ms / 4093 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 310 ms / 3763 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 312 ms / 3763 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 320 ms / 3763 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 321 ms / 4093 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 326 ms / 4093 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 332 ms / 4335 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 351 ms / 3763 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 364 ms / 4093 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 377 ms / 4093 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 462 ms / 4093 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 463 ms / 4093 ms |
FP16 | mps | Failed | — | macos · aarch64 | 2100 ms / 2555 ms |
BF16 | mps | Failed | — | macos · aarch64 | 2000 ms / 2632 ms |
Test setup & sample details
- Quality gate
- { "minimumTop1Accuracy": 1, "maximumCpuScoreDifference": 0.03, "finiteScoresInUnitInterval": true, "scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification", "performanceTargets": { "goodInferenceMs": 250, "acceptableInferenceMs": 1000, "description": "One image at pinned256px FixRes or256patch NaFlex default and four labels; good<=250ms,acceptable<=1000ms" }, "version": 2, "minimumLogOddsMargin": 1, "gateRevisionReason": "V1 required0.05absolute sigmoid margin, which conflated calibrated confidence with ranking despite the native independent sigmoid contract. All six semantic fixtures ranked correctly but dog/portrait failed absolute margin. Preserve fullv1failures;v2requires expected top1plus ≥1natural-log-odds separation (odds ratio≥e), finite scores and unchanged0.03CPU agreement. This measures ranking separation without treating sigmoid scores as calibrated probabilities." }
- Tested
- 2026-11-06T17:19:37.653679+00:00
- Environment
- { "hardware": "Apple M3 Max", "os": "macOS-27.2-arm64-arm-64bit-Mach-O", "runtime": "Transformers5.22.1/Torch2.10.0/Pillow12.0.0", "interpreterScope": "Existing isolated private probe venv from app-ownedCPython3.13.18; managed packaged provision is separate evidence" }
- Runtime
- transformers 5.22.1
- Artifact revision
- 3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab
- Fixture
{ "file": "model-samples/desktop/siglip2-family/siglip2-base-patch16-256-dog-input.json", "size": 567, "sha1": "f9b27f8c462739815e725c2824899b4eb249c8be", "sha256": "d280bd6cb58649d1f878f8ec5d188b9ff0b5feb0fd2127e374996c7fa9f3b7c8", "url": "https://assets.localai.tools/samples/v1/files/d280bd6cb58649d1f878f8ec5d188b9ff0b5feb0fd2127e374996c7fa9f3b7c8/siglip2-base-patch16-256-dog-input.json" }- Rerun when
- Original weights/processor/tokenizer artifact revision; Pinned runtime or relevant OS/driver/backend behavior; Fixture bytes/labels/expected result or quality gate; Image preprocessing/prompt template/sigmoid scoring
Per-case quality & failure details
[
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2546.8485000019427,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 246.39604199910536,
"inferenceMs": 226.81179199935286,
"imagesPerSecond": 4.408941841978186,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2546.8485000019427,
"totalMs": 2793.244542001048,
"initializationReused": false,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450852352
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00863016560106189,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.42086227124568,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2546.8485000019427,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 173.37508400305524,
"inferenceMs": 161.93145899887895,
"imagesPerSecond": 6.175452294337217,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2546.8485000019427,
"totalMs": 2720.223584004998,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450852352
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0008411059180275515,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.188875144637824,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2546.8485000019427,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 175.22995800027275,
"inferenceMs": 161.8372500015539,
"imagesPerSecond": 6.179047159973358,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2546.8485000019427,
"totalMs": 2722.0784580022155,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450852352
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0013876414509468304,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.126726092506584,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2546.8485000019427,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 173.4134170001198,
"inferenceMs": 160.59108400077093,
"imagesPerSecond": 6.2269957651895504,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2546.8485000019427,
"totalMs": 2720.2619170020625,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450852352
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.2790693828537769,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.057682043277175,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2546.8485000019427,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 166.93445800046902,
"inferenceMs": 158.85158400124055,
"imagesPerSecond": 6.2951843148897435,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2546.8485000019427,
"totalMs": 2713.7829580024118,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450852352
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.28660052852501394,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.988556948880861,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2546.8485000019427,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 173.47054199854028,
"inferenceMs": 165.62370800238568,
"imagesPerSecond": 6.037782948233448,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2546.8485000019427,
"totalMs": 2720.319042000483,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450852352
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0009220244828611612,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.4660921464207641,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2546.8485000019427,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 179.01324999911594,
"inferenceMs": 175.95437500131084,
"imagesPerSecond": 5.683291478217295,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2546.8485000019427,
"totalMs": 2725.8617500010587,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450852352
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.02168915234688029,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.650545119013157,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2546.8485000019427,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 163.2962090006913,
"inferenceMs": 158.56708399951458,
"imagesPerSecond": 6.30647909248972,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2546.8485000019427,
"totalMs": 2710.144709002634,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450852352
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.004866089934694173,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.048371290918458,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2546.8485000019427,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 162.64229100124794,
"inferenceMs": 159.49545900002704,
"imagesPerSecond": 6.269770978243527,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2546.8485000019427,
"totalMs": 2709.4907910031907,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450852352
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2546.8485000019427,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 168.1913339998573,
"inferenceMs": 157.17745799702243,
"imagesPerSecond": 6.362235480478021,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2715.0398340018,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450852352
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2546.8485000019427,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 169.25854199871537,
"inferenceMs": 158.3102090007742,
"imagesPerSecond": 6.316712019470012,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2716.107042000658,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450852352
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2546.8485000019427,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 184.09845800124458,
"inferenceMs": 172.61683299875585,
"imagesPerSecond": 5.793177771991725,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2730.9469580031873,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450852352
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2546.8485000019427,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 167.98266700061504,
"inferenceMs": 156.7727500005276,
"imagesPerSecond": 6.3786595565660145,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2714.831167002558,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450852352
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2546.8485000019427,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 170.8239590006997,
"inferenceMs": 159.8740830013412,
"imagesPerSecond": 6.254922506680528,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2717.6724590026424,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450852352
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2431.714665999607,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 237.46641599791474,
"inferenceMs": 218.64216700123507,
"imagesPerSecond": 4.573683172442904,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2431.714665999607,
"totalMs": 2669.1810819975217,
"initializationReused": false,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1449132032
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00863016560106189,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.42086227124568,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2431.714665999607,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 171.0322089966212,
"inferenceMs": 159.71533300034935,
"imagesPerSecond": 6.261139623944638,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2431.714665999607,
"totalMs": 2602.746874996228,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1449132032
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0008411059180275515,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.188875144637824,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2431.714665999607,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 173.6060000002908,
"inferenceMs": 160.34141699856264,
"imagesPerSecond": 6.236691796287197,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2431.714665999607,
"totalMs": 2605.3206659998978,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1449132032
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0013876414509468304,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.126726092506584,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2431.714665999607,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 171.4035420009168,
"inferenceMs": 158.85504100151593,
"imagesPerSecond": 6.295047319212597,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2431.714665999607,
"totalMs": 2603.1182080005237,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1449132032
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.2790693828537769,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.057682043277175,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2431.714665999607,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 168.29462500027148,
"inferenceMs": 160.22937500019907,
"imagesPerSecond": 6.241052865610676,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2431.714665999607,
"totalMs": 2600.0092909998784,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1449132032
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.28660052852501394,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.988556948880861,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2431.714665999607,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 167.00637499889126,
"inferenceMs": 158.992999997281,
"imagesPerSecond": 6.289585076180092,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2431.714665999607,
"totalMs": 2598.721040998498,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1449132032
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0009220244828611612,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.4660921464207641,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2431.714665999607,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 166.57658300027833,
"inferenceMs": 164.14241700113053,
"imagesPerSecond": 6.0922704701802495,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2431.714665999607,
"totalMs": 2598.2912489998853,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1449132032
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.02168915234688029,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.650545119013157,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2431.714665999607,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 167.07737500109943,
"inferenceMs": 162.64283299824456,
"imagesPerSecond": 6.14844184379642,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2431.714665999607,
"totalMs": 2598.7920410007064,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1449132032
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.004866089934694173,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.048371290918458,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2431.714665999607,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 162.7815419997205,
"inferenceMs": 159.58741700160317,
"imagesPerSecond": 6.26615818958931,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2431.714665999607,
"totalMs": 2594.4962079993275,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1449132032
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2431.714665999607,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 169.76441699807765,
"inferenceMs": 158.84987500248826,
"imagesPerSecond": 6.295252042120498,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2601.4790829976846,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1449132032
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2431.714665999607,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 171.02954200163367,
"inferenceMs": 159.8999170018942,
"imagesPerSecond": 6.25391193910472,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2602.7442080012406,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1449132032
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2431.714665999607,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 168.64950000308454,
"inferenceMs": 157.34683300252073,
"imagesPerSecond": 6.355386892241929,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2600.3641660026915,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1449132032
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2431.714665999607,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 166.27745899677393,
"inferenceMs": 155.07541700208094,
"imagesPerSecond": 6.448475324664651,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2597.992124996381,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1449132032
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2431.714665999607,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 168.3069999999134,
"inferenceMs": 156.73079100088216,
"imagesPerSecond": 6.380367211917992,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2600.0216659995203,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1449132032
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 3763.308916000824,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 302.6409579979372,
"inferenceMs": 279.6815000001516,
"imagesPerSecond": 3.575495697782864,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 3763.308916000824,
"totalMs": 4065.949873998761,
"initializationReused": false,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1454292992
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00863016560106189,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.42086227124568,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 3763.308916000824,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 370.39687499782303,
"inferenceMs": 351.33425000094576,
"imagesPerSecond": 2.846292383954334,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 3763.308916000824,
"totalMs": 4133.705790998647,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1454292992
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0008411059180275515,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.188875144637824,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 3763.308916000824,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 328.5096250001516,
"inferenceMs": 312.2734170028707,
"imagesPerSecond": 3.202321893415625,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 3763.308916000824,
"totalMs": 4091.8185410009755,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1454292992
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0013876414509468304,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.126726092506584,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 3763.308916000824,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 180.22520799786435,
"inferenceMs": 166.66616599832196,
"imagesPerSecond": 6.000018024114554,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 3763.308916000824,
"totalMs": 3943.5341239986883,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454292992
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.2790693828537769,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.057682043277175,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 3763.308916000824,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 165.39212499992573,
"inferenceMs": 157.10320899961516,
"imagesPerSecond": 6.365242354804157,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 3763.308916000824,
"totalMs": 3928.7010410007497,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454292992
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.28660052852501394,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.988556948880861,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 3763.308916000824,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 164.25066699957824,
"inferenceMs": 156.3316659994598,
"imagesPerSecond": 6.3966567080750965,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 3763.308916000824,
"totalMs": 3927.559583000402,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454292992
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0009220244828611612,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.4660921464207641,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 3763.308916000824,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 198.6943749980128,
"inferenceMs": 196.31445800041547,
"imagesPerSecond": 5.0938683283219195,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 3763.308916000824,
"totalMs": 3962.0032909988367,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454292992
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.02168915234688029,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.650545119013157,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 3763.308916000824,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 176.50966700239223,
"inferenceMs": 169.37195799982874,
"imagesPerSecond": 5.904165080272681,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 3763.308916000824,
"totalMs": 3939.818583003216,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454292992
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.004866089934694173,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.048371290918458,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 3763.308916000824,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 174.8741669980518,
"inferenceMs": 171.72875000323984,
"imagesPerSecond": 5.823136778094139,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 3763.308916000824,
"totalMs": 3938.1830829988758,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454292992
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 3763.308916000824,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 345.7476250005129,
"inferenceMs": 309.74649999916437,
"imagesPerSecond": 3.2284464877010643,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 4109.056541001337,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1454292992
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 3763.308916000824,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 331.6668330007815,
"inferenceMs": 320.24204199842643,
"imagesPerSecond": 3.1226380951096786,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 4094.9757490016054,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1454292992
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 3763.308916000824,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 226.39699999854201,
"inferenceMs": 214.01191699987976,
"imagesPerSecond": 4.672636991521187,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 3989.705915999366,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454292992
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 3763.308916000824,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 277.7411250026489,
"inferenceMs": 261.42266700117034,
"imagesPerSecond": 3.8252230056069436,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 4041.050041003473,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1454292992
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 3763.308916000824,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 300.37875000198255,
"inferenceMs": 280.2579589988454,
"imagesPerSecond": 3.5681413065743466,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 4063.6876660028065,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1454292992
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4092.675957999745,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 485.11904099723324,
"inferenceMs": 462.5004999979865,
"imagesPerSecond": 2.1621598247014946,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 4092.675957999745,
"totalMs": 4577.794998996978,
"initializationReused": false,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1460584448
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00863016560106189,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.42086227124568,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4092.675957999745,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 263.91837499977555,
"inferenceMs": 249.41770800069207,
"imagesPerSecond": 4.009338422744327,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 4092.675957999745,
"totalMs": 4356.594332999521,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1460584448
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0008411059180275515,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.188875144637824,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4092.675957999745,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 305.0256249989616,
"inferenceMs": 286.7246249988966,
"imagesPerSecond": 3.487666955720487,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 4092.675957999745,
"totalMs": 4397.701582998707,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1460584448
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0013876414509468304,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.126726092506584,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4092.675957999745,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 351.97216700180434,
"inferenceMs": 325.9353749999718,
"imagesPerSecond": 3.0680928696373826,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 4092.675957999745,
"totalMs": 4444.6481250015495,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1460584448
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.2790693828537769,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.057682043277175,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4092.675957999745,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 374.046291999548,
"inferenceMs": 363.93870799656725,
"imagesPerSecond": 2.7477154202828906,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 4092.675957999745,
"totalMs": 4466.722249999293,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1460584448
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.28660052852501394,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.988556948880861,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4092.675957999745,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 481.08808300094097,
"inferenceMs": 462.0255419977184,
"imagesPerSecond": 2.1643825050800727,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 4092.675957999745,
"totalMs": 4573.764041000686,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1460584448
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0009220244828611612,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.4660921464207641,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4092.675957999745,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 235.18491700087907,
"inferenceMs": 231.499082998198,
"imagesPerSecond": 4.319671538430172,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 4092.675957999745,
"totalMs": 4327.860875000624,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1460584448
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.02168915234688029,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.650545119013157,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4092.675957999745,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 386.86970900016604,
"inferenceMs": 377.13041700044414,
"imagesPerSecond": 2.651602615226929,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 4092.675957999745,
"totalMs": 4479.545666999911,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1460584448
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.004866089934694173,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.048371290918458,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4092.675957999745,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 326.08416700168164,
"inferenceMs": 320.71349999750964,
"imagesPerSecond": 3.1180477279807834,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 4092.675957999745,
"totalMs": 4418.760125001427,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1460584448
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4092.675957999745,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 229.4815409986768,
"inferenceMs": 214.50745900074253,
"imagesPerSecond": 4.661842551575507,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 4322.157498998422,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1460584448
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4092.675957999745,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 296.12687500048196,
"inferenceMs": 276.45599999959813,
"imagesPerSecond": 3.6172121422629773,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 4388.802833000227,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1460584448
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4092.675957999745,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 226.63120800280012,
"inferenceMs": 203.74416599952383,
"imagesPerSecond": 4.9081159948517845,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 4319.307166002545,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1460584448
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4092.675957999745,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 225.70400000040536,
"inferenceMs": 209.93137500045123,
"imagesPerSecond": 4.76346139302832,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 4318.3799580001505,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1460584448
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4092.675957999745,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 246.3608749967534,
"inferenceMs": 231.77579099865397,
"imagesPerSecond": 4.31451445248571,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 4339.036832996499,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1460584448
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4335.38304099784,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 369.5617500015942,
"inferenceMs": 331.5125410008477,
"imagesPerSecond": 3.016477135317312,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 4335.38304099784,
"totalMs": 4704.944790999434,
"initializationReused": false,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1456979968
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00863016560106189,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.42086227124568,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4335.38304099784,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 273.5717499999737,
"inferenceMs": 259.2285830032779,
"imagesPerSecond": 3.857599298713735,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 4335.38304099784,
"totalMs": 4608.954790997814,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1456979968
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0008411059180275515,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.188875144637824,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4335.38304099784,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 283.69225000278675,
"inferenceMs": 262.66554099856876,
"imagesPerSecond": 3.8071229145563823,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 4335.38304099784,
"totalMs": 4619.075291000627,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1456979968
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0013876414509468304,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.126726092506584,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4335.38304099784,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 237.51929200079758,
"inferenceMs": 207.26195899987943,
"imagesPerSecond": 4.8248120630789835,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 4335.38304099784,
"totalMs": 4572.902332998638,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456979968
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.2790693828537769,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.057682043277175,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4335.38304099784,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 242.0122920011636,
"inferenceMs": 232.53291600121884,
"imagesPerSecond": 4.300466433727423,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 4335.38304099784,
"totalMs": 4577.395332999004,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456979968
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.28660052852501394,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.988556948880861,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4335.38304099784,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 228.53583400137722,
"inferenceMs": 219.37404200070887,
"imagesPerSecond": 4.558424464808688,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 4335.38304099784,
"totalMs": 4563.918874999217,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456979968
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0009220244828611612,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.4660921464207641,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4335.38304099784,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 185.33370799923432,
"inferenceMs": 182.4426669991226,
"imagesPerSecond": 5.4811739843992155,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 4335.38304099784,
"totalMs": 4520.716748997074,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456979968
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.02168915234688029,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.650545119013157,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4335.38304099784,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 127.08237500191899,
"inferenceMs": 122.0904999972845,
"imagesPerSecond": 8.19064546399795,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 4335.38304099784,
"totalMs": 4462.465415999759,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456979968
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.004866089934694173,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.048371290918458,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4335.38304099784,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 101.48266599935596,
"inferenceMs": 98.01849999712431,
"imagesPerSecond": 10.202155715801998,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 4335.38304099784,
"totalMs": 4436.865706997196,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456979968
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4335.38304099784,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 109.26491699865437,
"inferenceMs": 95.86112499891897,
"imagesPerSecond": 10.431757399167568,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 4444.647957996494,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456979968
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4335.38304099784,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 112.64229199878173,
"inferenceMs": 101.41183400264708,
"imagesPerSecond": 9.860782125031855,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 4448.025332996622,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456979968
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4335.38304099784,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 117.42162500013364,
"inferenceMs": 106.02408300110255,
"imagesPerSecond": 9.431819372487295,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 4452.804665997974,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456979968
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4335.38304099784,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 108.50845800086972,
"inferenceMs": 96.1854580018553,
"imagesPerSecond": 10.396581986236539,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 4443.89149899871,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456979968
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 4335.38304099784,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 110.1087089991779,
"inferenceMs": 98.62404099840205,
"imagesPerSecond": 10.139515577304346,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 4445.491749997018,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456979968
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.186292000144,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 173.11504199824412,
"inferenceMs": 153.48795799945947,
"imagesPerSecond": 6.515169092310953,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2430.186292000144,
"totalMs": 2603.301333998388,
"initializationReused": false,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450655744
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00863016560106189,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.42086227124568,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.186292000144,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 100.12229100175318,
"inferenceMs": 89.08895900094649,
"imagesPerSecond": 11.224735491514453,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2430.186292000144,
"totalMs": 2530.308583001897,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450655744
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0008411059180275515,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.188875144637824,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.186292000144,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 101.5730840008473,
"inferenceMs": 88.82291600093595,
"imagesPerSecond": 11.258355895335194,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2430.186292000144,
"totalMs": 2531.7593760009913,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450655744
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0013876414509468304,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.126726092506584,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.186292000144,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 101.8274999987625,
"inferenceMs": 89.54004200131749,
"imagesPerSecond": 11.168187747614482,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2430.186292000144,
"totalMs": 2532.0137919989065,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450655744
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.2790693828537769,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.057682043277175,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.186292000144,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 97.60487500170711,
"inferenceMs": 89.80070899997372,
"imagesPerSecond": 11.13576954053105,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2430.186292000144,
"totalMs": 2527.791167001851,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450655744
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.28660052852501394,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.988556948880861,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.186292000144,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 97.62799999953131,
"inferenceMs": 90.03595900139771,
"imagesPerSecond": 11.106673501244941,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2430.186292000144,
"totalMs": 2527.8142919996753,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450655744
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0009220244828611612,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.4660921464207641,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.186292000144,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 93.35879100035527,
"inferenceMs": 91.05616699889651,
"imagesPerSecond": 10.982232537990742,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2430.186292000144,
"totalMs": 2523.5450830004993,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450655744
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.02168915234688029,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.650545119013157,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.186292000144,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 95.25629099880462,
"inferenceMs": 91.28454199890257,
"imagesPerSecond": 10.954757268892493,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2430.186292000144,
"totalMs": 2525.4425829989486,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450655744
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.004866089934694173,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.048371290918458,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.186292000144,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 96.66424999886658,
"inferenceMs": 93.68866600198089,
"imagesPerSecond": 10.673649681156274,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2430.186292000144,
"totalMs": 2526.8505419990106,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450655744
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.186292000144,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 100.21070800212328,
"inferenceMs": 89.54062500197324,
"imagesPerSecond": 11.16811503133871,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2530.3970000022673,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450655744
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.186292000144,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 100.7086249992426,
"inferenceMs": 90.03333399959956,
"imagesPerSecond": 11.106997326173078,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2530.8949169993866,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450655744
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.186292000144,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 100.12566599834827,
"inferenceMs": 89.39037500022096,
"imagesPerSecond": 11.186886731345831,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2530.3119579984923,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450655744
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.186292000144,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 100.31941600027494,
"inferenceMs": 89.61374999853433,
"imagesPerSecond": 11.159001827469059,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2530.505708000419,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450655744
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.00993366652255645,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.606259343222257,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.186292000144,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 101.3171250015148,
"inferenceMs": 90.75904099881882,
"imagesPerSecond": 11.018186056119902,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2531.503417001659,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1450655744
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 163.22924999985844,
"inferenceMs": 131.17816700105323,
"imagesPerSecond": 7.623219799922734,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571094528,
"processInitMs": 2696.145833000628,
"totalMs": 2859.3750830004865,
"initializationReused": false,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.940696716308594e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.008630254931631498,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.420870699734089,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.1175839973439,
"inferenceMs": 42.33962500074995,
"imagesPerSecond": 23.618537008352988,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571094528,
"processInitMs": 2696.145833000628,
"totalMs": 2751.263416997972,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0419171303510666e-8,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0008411163421442325,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.188895308909132,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.939875000243774,
"inferenceMs": 36.97120899960282,
"imagesPerSecond": 27.04807408409995,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571094528,
"processInitMs": 2696.145833000628,
"totalMs": 2748.085708000872,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.280568540096283e-9,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.001387642752888496,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.126746147015463,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.81733399978839,
"inferenceMs": 42.18287499679718,
"imagesPerSecond": 23.706302618679427,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571094528,
"processInitMs": 2696.145833000628,
"totalMs": 2752.9631670004164,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 3.8743019104003906e-7,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.2790697698110307,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.057673482400045,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 52.499958997941576,
"inferenceMs": 42.190415999357356,
"imagesPerSecond": 23.702065417303114,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571094528,
"processInitMs": 2696.145833000628,
"totalMs": 2748.6457919985696,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 4.3291947804391384e-10,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.28660052809209446,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.988548317530151,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.71366699869395,
"inferenceMs": 41.82670799855259,
"imagesPerSecond": 23.90816891529223,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571094528,
"processInitMs": 2696.145833000628,
"totalMs": 2747.859499999322,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 4.6566128730773926e-9,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0009220190404448658,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.4660854208377616,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.79595899907872,
"inferenceMs": 26.63025000219932,
"imagesPerSecond": 37.551280964970765,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"processInitMs": 2696.145833000628,
"totalMs": 2726.9417919997068,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 2.0489096641540527e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.02168935704867181,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.650548173476675,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.55054200012819,
"inferenceMs": 25.505416000669356,
"imagesPerSecond": 39.207358937950914,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"processInitMs": 2696.145833000628,
"totalMs": 2726.6963750007562,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0151416063308716e-7,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0048661914165677445,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.048384642937736,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.781999997998355,
"inferenceMs": 25.996415999543387,
"imagesPerSecond": 38.46684096829211,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"processInitMs": 2696.145833000628,
"totalMs": 2726.9278329986264,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.013249998912215,
"inferenceMs": 40.80850000173086,
"imagesPerSecond": 24.504698774950946,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2749.1590829995403,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 59.28570800097077,
"inferenceMs": 47.11841699827346,
"imagesPerSecond": 21.223123859119514,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2755.431541001599,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.59658299898729,
"inferenceMs": 42.36933400170528,
"imagesPerSecond": 23.60197589982774,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2751.7424159996153,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 58.207124999171356,
"inferenceMs": 44.74900000059279,
"imagesPerSecond": 22.346868086141658,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2754.3529579997994,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 57.99762500100769,
"inferenceMs": 44.24541600019438,
"imagesPerSecond": 22.601211388669206,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2754.1434580016357,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.32879200129537,
"inferenceMs": 44.11633299969253,
"imagesPerSecond": 22.667341821156565,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2752.4746250019234,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 128.91700000182027,
"inferenceMs": 116.03829200248583,
"imagesPerSecond": 8.6178448746779,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2825.0628330024483,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 113.64816599962069,
"inferenceMs": 99.76787500272621,
"imagesPerSecond": 10.023266507106365,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2809.7939990002487,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 122.27641699792002,
"inferenceMs": 109.34879200067371,
"imagesPerSecond": 9.145048442728465,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2818.422249998548,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 114.66929100060952,
"inferenceMs": 102.31758400186663,
"imagesPerSecond": 9.773491133075977,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2810.8151240012376,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 121.95129099927726,
"inferenceMs": 108.61216600096668,
"imagesPerSecond": 9.207071701259505,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2818.0971239999053,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 125.86150000061025,
"inferenceMs": 110.51833300007274,
"imagesPerSecond": 9.04827256125318,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2822.0073330012383,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 126.23175000044284,
"inferenceMs": 113.27091600105632,
"imagesPerSecond": 8.82839157044227,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2822.377583001071,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 118.71524999878602,
"inferenceMs": 105.55875000136439,
"imagesPerSecond": 9.473397515479054,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2814.861082999414,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 117.7940830020816,
"inferenceMs": 103.91774999879999,
"imagesPerSecond": 9.62299510922386,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2813.9399160027097,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 121.28750000192667,
"inferenceMs": 105.38950000045588,
"imagesPerSecond": 9.488611294252978,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2817.4333330025547,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 118.31116699977429,
"inferenceMs": 102.07199999786098,
"imagesPerSecond": 9.797006035161022,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2814.4570000004023,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 124.90316700132098,
"inferenceMs": 112.079000002268,
"imagesPerSecond": 8.92227803584761,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2821.049000001949,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 125.16391699682572,
"inferenceMs": 110.54158299884875,
"imagesPerSecond": 9.046369455469211,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2821.3097499974538,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 111.6755840012047,
"inferenceMs": 94.915999998193,
"imagesPerSecond": 10.535631505953031,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2807.8214170018327,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 111.90583300049184,
"inferenceMs": 98.99491599935573,
"imagesPerSecond": 10.101528850294779,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2808.05166600112,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 71.21050000205287,
"inferenceMs": 58.69287500172504,
"imagesPerSecond": 17.037843178931155,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2767.356333002681,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 61.555666998174274,
"inferenceMs": 48.615208001137944,
"imagesPerSecond": 20.569694980562314,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2757.7014999988023,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.482040998962475,
"inferenceMs": 38.64604200134636,
"imagesPerSecond": 25.87587106501519,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2747.6278739995905,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.163334000098985,
"inferenceMs": 42.494500001339475,
"imagesPerSecond": 23.53245714077066,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2751.309167000727,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.8288330026262,
"inferenceMs": 41.419417000724934,
"imagesPerSecond": 24.14326594656071,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2749.9746660032542,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.54195800141315,
"inferenceMs": 42.734332997497404,
"imagesPerSecond": 23.40038863034464,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2751.687791002041,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.89954199967906,
"inferenceMs": 43.254457999864826,
"imagesPerSecond": 23.119004288601307,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2752.045375000307,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 60.163874997670064,
"inferenceMs": 47.249582999938866,
"imagesPerSecond": 21.16420794658217,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2756.309707998298,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2696.145833000628,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.87529200076824,
"inferenceMs": 44.087917001888854,
"imagesPerSecond": 22.681951609488767,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2753.0211250013963,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 942112768
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2805.509792000521,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 163.59970800112933,
"inferenceMs": 131.0035410024284,
"imagesPerSecond": 7.633381451738493,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571094528,
"processInitMs": 2805.509792000521,
"totalMs": 2969.1095000016503,
"initializationReused": false,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 800030720
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.940696716308594e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.008630254931631498,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.420870699734089,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2805.509792000521,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.07883299974492,
"inferenceMs": 39.64554099729867,
"imagesPerSecond": 25.22351757208048,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571094528,
"processInitMs": 2805.509792000521,
"totalMs": 2858.588625000266,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 800030720
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0419171303510666e-8,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0008411163421442325,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.188895308909132,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2805.509792000521,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 59.2330840008799,
"inferenceMs": 44.03833300239057,
"imagesPerSecond": 22.707489857659148,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571094528,
"processInitMs": 2805.509792000521,
"totalMs": 2864.742876001401,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 800030720
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.280568540096283e-9,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.001387642752888496,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.126746147015463,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2805.509792000521,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 50.632124999538064,
"inferenceMs": 36.6230830004497,
"imagesPerSecond": 27.30518345459122,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571094528,
"processInitMs": 2805.509792000521,
"totalMs": 2856.141917000059,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 800030720
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 3.8743019104003906e-7,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.2790697698110307,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.057673482400045,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2805.509792000521,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 49.29404200083809,
"inferenceMs": 38.89133399934508,
"imagesPerSecond": 25.712669049018473,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571094528,
"processInitMs": 2805.509792000521,
"totalMs": 2854.803834001359,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 800030720
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 4.3291947804391384e-10,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.28660052809209446,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.988548317530151,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2805.509792000521,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 45.437124997988576,
"inferenceMs": 36.038707999978215,
"imagesPerSecond": 27.74794257331879,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571094528,
"processInitMs": 2805.509792000521,
"totalMs": 2850.9469169985096,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 800030720
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 4.6566128730773926e-9,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0009220190404448658,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.4660854208377616,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2805.509792000521,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.331417001434602,
"inferenceMs": 25.23141599886003,
"imagesPerSecond": 39.63313038178993,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"processInitMs": 2805.509792000521,
"totalMs": 2835.8412090019556,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 800030720
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 2.0489096641540527e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.02168935704867181,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.650548173476675,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2805.509792000521,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 32.50000000116415,
"inferenceMs": 26.43175000048359,
"imagesPerSecond": 37.833287617418605,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"processInitMs": 2805.509792000521,
"totalMs": 2838.009792001685,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 800030720
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0151416063308716e-7,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0048661914165677445,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.048384642937736,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2805.509792000521,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 29.66749999905005,
"inferenceMs": 24.955832999694394,
"imagesPerSecond": 40.07079226777346,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"processInitMs": 2805.509792000521,
"totalMs": 2835.177291999571,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 800030720
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2805.509792000521,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.147999999666354,
"inferenceMs": 43.33208400203148,
"imagesPerSecond": 23.077588420467347,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2861.6577920001873,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 800030720
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2805.509792000521,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 57.404916999075795,
"inferenceMs": 43.68408300069859,
"imagesPerSecond": 22.89163309171462,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2862.9147089995968,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 800030720
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2805.509792000521,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 58.79174999790848,
"inferenceMs": 45.525416997406865,
"imagesPerSecond": 21.965751572510804,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2864.3015419984295,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 800030720
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2805.509792000521,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.49758299678797,
"inferenceMs": 34.65404099915759,
"imagesPerSecond": 28.856663499195065,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2854.007374997309,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 800030720
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2805.509792000521,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 64.42116700054612,
"inferenceMs": 51.268458999402355,
"imagesPerSecond": 19.505169835739693,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2869.930959001067,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 800030720
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2695.804291000968,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 163.13704200001666,
"inferenceMs": 130.98429099773057,
"imagesPerSecond": 7.6345032857209265,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571094528,
"processInitMs": 2695.804291000968,
"totalMs": 2858.9413330009847,
"initializationReused": false,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 797343744
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.940696716308594e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.008630254931631498,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.420870699734089,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2695.804291000968,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 58.59295800109976,
"inferenceMs": 45.60175000005984,
"imagesPerSecond": 21.928982988562673,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571094528,
"processInitMs": 2695.804291000968,
"totalMs": 2754.397249002068,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797343744
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0419171303510666e-8,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0008411163421442325,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.188895308909132,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2695.804291000968,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 59.45541599794524,
"inferenceMs": 45.06787499849452,
"imagesPerSecond": 22.18875418540157,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571094528,
"processInitMs": 2695.804291000968,
"totalMs": 2755.2597069989133,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797343744
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.280568540096283e-9,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.001387642752888496,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.126746147015463,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2695.804291000968,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 58.04345900105545,
"inferenceMs": 43.35362499841722,
"imagesPerSecond": 23.06612192259606,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571094528,
"processInitMs": 2695.804291000968,
"totalMs": 2753.8477500020235,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797343744
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 3.8743019104003906e-7,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.2790697698110307,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.057673482400045,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2695.804291000968,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.70958400089876,
"inferenceMs": 43.79970899753971,
"imagesPerSecond": 22.83120191634541,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571094528,
"processInitMs": 2695.804291000968,
"totalMs": 2749.513875001867,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797343744
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 4.3291947804391384e-10,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.28660052809209446,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.988548317530151,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2695.804291000968,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 52.11866700119572,
"inferenceMs": 42.79733300063526,
"imagesPerSecond": 23.365941984869863,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"processInitMs": 2695.804291000968,
"totalMs": 2747.922958002164,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797343744
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 4.6566128730773926e-9,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0009220190404448658,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.4660854208377616,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2695.804291000968,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 29.784709000523435,
"inferenceMs": 25.30704200034961,
"imagesPerSecond": 39.514693182481984,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"processInitMs": 2695.804291000968,
"totalMs": 2725.5890000014915,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797343744
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 2.0489096641540527e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.02168935704867181,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.650548173476675,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2695.804291000968,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 32.66170900315046,
"inferenceMs": 26.310208999348106,
"imagesPerSecond": 38.0080599141108,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"processInitMs": 2695.804291000968,
"totalMs": 2728.4660000041185,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797343744
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0151416063308716e-7,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0048661914165677445,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.048384642937736,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2695.804291000968,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.877459001203533,
"inferenceMs": 25.659957998868776,
"imagesPerSecond": 38.97122512998989,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"processInitMs": 2695.804291000968,
"totalMs": 2726.6817500021716,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797343744
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2695.804291000968,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 57.518417001119815,
"inferenceMs": 45.2535420008644,
"imagesPerSecond": 22.097717787060706,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2753.322708002088,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797343744
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2695.804291000968,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.938083001819905,
"inferenceMs": 43.241375002253335,
"imagesPerSecond": 23.125999114225422,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2751.742374002788,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797343744
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2695.804291000968,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 49.36295800143853,
"inferenceMs": 35.707999999431195,
"imagesPerSecond": 28.004928867926775,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2745.1672490024066,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797343744
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2695.804291000968,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.98370900002192,
"inferenceMs": 40.88729199793306,
"imagesPerSecond": 24.457476911177,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2749.78800000099,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797343744
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.955777406692505e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009933647284924518,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.60626601296723,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2695.804291000968,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 62.0600000002014,
"inferenceMs": 48.900542002229486,
"imagesPerSecond": 20.449671088602816,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1571078144,
"totalMs": 2757.8642910011695,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797343744
}
},
{
"precision": "fp16",
"device": "mps",
"status": "fail",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00012064166367053986,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010053679510747315,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.601562626859118,
"passed": false
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2554.7815409991017,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 2130.4328340011125,
"inferenceMs": 2099.795374997484,
"imagesPerSecond": 0.47623688093950495,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113899008,
"processInitMs": 2554.7815409991017,
"totalMs": 4685.214375000214,
"initializationReused": false,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "slow",
"observedResidentBytesAtCaseEnd": 1505689600
}
},
{
"precision": "bf16",
"device": "mps",
"status": "fail",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0010162675753235817,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010946878173854202,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.624999917376456,
"passed": false
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2632.3861250020855,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 2033.1797920007375,
"inferenceMs": 2000.4606250004144,
"imagesPerSecond": 0.4998848702657133,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113899008,
"processInitMs": 2632.3861250020855,
"totalMs": 4665.565917002823,
"initializationReused": false,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "3f9f96cb90da5dbc758b01813f2f6f1aee24c1ab",
"checkpoint": "siglip2-base-patch16-256",
"fallback": false,
"grade": "slow",
"observedResidentBytesAtCaseEnd": 1406910464
}
}
]SigLIP 2 Base — Image & Text Matching (384) Desktop
145 configurations testedQuality criteria
{ "minimumTop1Accuracy": 1, "maximumCpuScoreDifference": 0.03, "finiteScoresInUnitInterval": true, "scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification", "performanceTargets": { "goodInferenceMs": 250, "acceptableInferenceMs": 1000, "description": "One384x384 image and four labels; good below250ms, acceptable below1s, slower cases retained but not recommended" }, "version": 2, "minimumLogOddsMargin": 1, "gateRevisionReason": "V1 required0.05absolute sigmoid margin, which conflated calibrated confidence with ranking despite the native independent sigmoid contract. All six semantic fixtures ranked correctly but dog/portrait failed absolute margin. Preserve fullv1failures;v2requires expected top1plus ≥1natural-log-odds separation (odds ratio≥e), finite scores and unchanged0.03CPU agreement. This measures ranking separation without treating sigmoid scores as calibrated probabilities." }
| Precision | Device | Result | Platform | Inference / load | |
|---|---|---|---|---|---|
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 94 ms / 825 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 100 ms / 885 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 100 ms / 859 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 104 ms / 835 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 124 ms / 922 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 132 ms / 887 ms |
BF16 | mpsAccelerated | Passed | Moderate | macos · aarch64 | 518 ms / 838 ms |
FP32 | mpsAccelerated | Passed | Moderate | macos · aarch64 | 527 ms / 929 ms |
FP16 | mpsAccelerated | Passed | Moderate | macos · aarch64 | 577 ms / 793 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 138 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 139 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 140 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 140 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 140 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 140 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 140 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 140 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 140 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 141 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 141 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 141 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 142 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 142 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 142 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 142 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 142 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 143 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 143 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 144 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 146 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 207 ms / 645 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 208 ms / 634 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 208 ms / 656 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 266 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 270 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 270 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 271 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 271 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 271 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 271 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 272 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 272 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 273 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 273 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 273 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 273 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 273 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 273 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 273 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 274 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 274 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 275 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 277 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 284 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 333 ms / 658 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 333 ms / 632 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 360 ms / 668 ms |
Test setup & sample details
- Quality gate
- { "minimumTop1Accuracy": 1, "maximumCpuScoreDifference": 0.03, "finiteScoresInUnitInterval": true, "scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification", "performanceTargets": { "goodInferenceMs": 250, "acceptableInferenceMs": 1000, "description": "One384x384 image and four labels; good below250ms, acceptable below1s, slower cases retained but not recommended" }, "version": 2, "minimumLogOddsMargin": 1, "gateRevisionReason": "V1 required0.05absolute sigmoid margin, which conflated calibrated confidence with ranking despite the native independent sigmoid contract. All six semantic fixtures ranked correctly but dog/portrait failed absolute margin. Preserve fullv1failures;v2requires expected top1plus ≥1natural-log-odds separation (odds ratio≥e), finite scores and unchanged0.03CPU agreement. This measures ranking separation without treating sigmoid scores as calibrated probabilities." }
- Tested
- 2026-11-02T17:30:07.061050+00:00
- Environment
- { "hardware": "Apple M3 Max", "os": "macOS-27.2-arm64-arm-64bit-Mach-O", "runtime": "Transformers5.22.1/Torch2.10.0/Pillow12.0.0", "interpreterScope": "Existing isolated private probe venv from app-ownedCPython3.13.18; managed packaged provision is separate evidence" }
- Runtime
- transformers 5.22.1
- Artifact revision
- f775b65a79762255128c981547af89addcfe0f88
- Fixture
{ "file": "model-samples/desktop/dinov2/dinov2-dog.png", "size": 709899, "sha1": "409baef01589c0a55c12bba3b4f5330adc74b1c7", "sha256": "cc986ccf120f7892c9ee0202001c30a91a3039cc5859d851b981476507a25fe3", "url": "https://assets.localai.tools/samples/v1/files/cc986ccf120f7892c9ee0202001c30a91a3039cc5859d851b981476507a25fe3/dinov2-dog.png" }- Rerun when
- Original weights/processor/tokenizer artifact revision; Pinned runtime or relevant OS/driver/backend behavior; Fixture bytes/labels/expected result or quality gate; Image preprocessing/prompt template/sigmoid scoring
Per-case quality & failure details
[
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 667.7769579982851,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 395.2537500008475,
"inferenceMs": 359.98437499802094,
"imagesPerSecond": 2.7778983462976625,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 3499.9217500007944,
"totalMs": 1063.0307079991326,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1482981376
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.012877492237748811,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.421337091234205,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 283.62483299861196,
"inferenceMs": 271.43700000306126,
"imagesPerSecond": 3.684096125394556,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 283.62483299861196,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1482981376
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0011582659442410659,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.08556940428876,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 286.3036249909783,
"inferenceMs": 272.8844579978613,
"imagesPerSecond": 3.664554615301093,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 286.3036249909783,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1482981376
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.001835596116279703,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.940511762680066,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 285.177041994757,
"inferenceMs": 270.9986249974463,
"imagesPerSecond": 3.690055623010719,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 285.177041994757,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1482981376
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.1671643669324112,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.287399368668465,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 292.9570410051383,
"inferenceMs": 283.76812499482185,
"imagesPerSecond": 3.5240039733787856,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 292.9570410051383,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1482981376
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.2348774507154303,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.268430687490877,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 281.93979199568275,
"inferenceMs": 272.7926670049783,
"imagesPerSecond": 3.665787687693785,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 281.93979199568275,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1482981376
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0011361937504261732,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.607579309107793,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 274.9998340004822,
"inferenceMs": 271.4527499920223,
"imagesPerSecond": 3.6838823700603105,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 274.9998340004822,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1482981376
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 285.3807909996249,
"inferenceMs": 273.2399170054123,
"imagesPerSecond": 3.659787380114715,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 285.3807909996249,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1482981376
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 285.1036250067409,
"inferenceMs": 273.2545000035316,
"imagesPerSecond": 3.659592065225187,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 285.1036250067409,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1482981376
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 284.3693339964375,
"inferenceMs": 272.68108299176674,
"imagesPerSecond": 3.667287767190633,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 284.3693339964375,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1482981376
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 285.05737500381656,
"inferenceMs": 272.60695800941903,
"imagesPerSecond": 3.668284945116655,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 285.05737500381656,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1482981376
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 286.7807080037892,
"inferenceMs": 274.9152909964323,
"imagesPerSecond": 3.6374841005587335,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 286.7807080037892,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1482981376
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 658.1933330016909,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 351.44766700977925,
"inferenceMs": 332.79920800123364,
"imagesPerSecond": 3.0048148431780315,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2165.7589170063147,
"totalMs": 1009.6410000114702,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1471709184
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 284.747833997244,
"inferenceMs": 273.3343749860069,
"imagesPerSecond": 3.658522643012589,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 284.747833997244,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1471709184
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 288.8935000082711,
"inferenceMs": 276.8846249964554,
"imagesPerSecond": 3.6116125986150434,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 288.8935000082711,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1471709184
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 286.4609170064796,
"inferenceMs": 274.0611670014914,
"imagesPerSecond": 3.648820483912477,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 286.4609170064796,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1471709184
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 285.9606660058489,
"inferenceMs": 274.02962499763817,
"imagesPerSecond": 3.649240479049004,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 285.9606660058489,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1471709184
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 281.8015000084415,
"inferenceMs": 270.0935000029858,
"imagesPerSecond": 3.702421568786162,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 281.8015000084415,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1471709184
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 632.3552500107326,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 351.77320800721645,
"inferenceMs": 333.03912499104626,
"imagesPerSecond": 3.002650214225986,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2106.4684580051107,
"totalMs": 984.128458017949,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1466908672
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 283.40524999657646,
"inferenceMs": 271.3966250012163,
"imagesPerSecond": 3.6846441992250947,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 283.40524999657646,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1466908672
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 283.18625000247266,
"inferenceMs": 271.6237089916831,
"imagesPerSecond": 3.6815637475542284,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 283.18625000247266,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1466908672
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 281.9667089934228,
"inferenceMs": 270.0269580091117,
"imagesPerSecond": 3.7033339462582706,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 281.9667089934228,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1466908672
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 283.8371669931803,
"inferenceMs": 271.8774170061806,
"imagesPerSecond": 3.6781282204739605,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 283.8371669931803,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1466908672
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 278.2112909917487,
"inferenceMs": 266.4195830002427,
"imagesPerSecond": 3.7534778364963097,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 278.2112909917487,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1466908672
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 634.0347079967614,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 227.22854198946152,
"inferenceMs": 207.67962500394788,
"imagesPerSecond": 4.8151088484534315,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2090.0017919921083,
"totalMs": 861.263249986223,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1489551360
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.012877492237748811,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.421337091234205,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 151.22641599737108,
"inferenceMs": 139.58945899503306,
"imagesPerSecond": 7.163864715856392,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 151.22641599737108,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1489551360
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0011582659442410659,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.08556940428876,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 153.90008299436886,
"inferenceMs": 140.45008399989456,
"imagesPerSecond": 7.119967261826278,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 153.90008299436886,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1489551360
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.001835596116279703,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.940511762680066,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 155.28904199891258,
"inferenceMs": 141.1309999966761,
"imagesPerSecond": 7.085615492156591,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 155.28904199891258,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1489551360
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.1671643669324112,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.287399368668465,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 151.44316699297633,
"inferenceMs": 142.55608299572486,
"imagesPerSecond": 7.014783087369125,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 151.44316699297633,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1489551360
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.2348774507154303,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.268430687490877,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 149.00108300207648,
"inferenceMs": 140.14575000328477,
"imagesPerSecond": 7.135428651789739,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 149.00108300207648,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1489551360
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0011361937504261732,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.607579309107793,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 141.9419169978937,
"inferenceMs": 138.4920830023475,
"imagesPerSecond": 7.22062935527549,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 141.9419169978937,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1489551360
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 158.5304170002928,
"inferenceMs": 146.0424170072656,
"imagesPerSecond": 6.847325732429161,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 158.5304170002928,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1489551360
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 154.58945799036883,
"inferenceMs": 142.42616599949542,
"imagesPerSecond": 7.021181767987371,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 154.58945799036883,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1489551360
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 151.91200000117533,
"inferenceMs": 140.23812500818167,
"imagesPerSecond": 7.130728537205262,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 151.91200000117533,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1489551360
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 151.0984170017764,
"inferenceMs": 139.24216601299122,
"imagesPerSecond": 7.181732578813091,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 151.0984170017764,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1489551360
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 151.60445800574962,
"inferenceMs": 139.84733300458174,
"imagesPerSecond": 7.150654778430687,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 151.60445800574962,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1489551360
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 645.2620000054594,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 225.97379199578427,
"inferenceMs": 206.89237500482704,
"imagesPerSecond": 4.833430908107023,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2160.0572080060374,
"totalMs": 871.2357920012437,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1465581568
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 153.49191600398626,
"inferenceMs": 141.7886660055956,
"imagesPerSecond": 7.052749900055733,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 153.49191600398626,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1465581568
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 153.47287499753293,
"inferenceMs": 141.56812500732485,
"imagesPerSecond": 7.063736981388001,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 153.47287499753293,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1465581568
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 152.56766699894797,
"inferenceMs": 140.4598340013763,
"imagesPerSecond": 7.119473030205927,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 152.56766699894797,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1465581568
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 151.71037500840612,
"inferenceMs": 139.71787500486244,
"imagesPerSecond": 7.157280340580604,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 151.71037500840612,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1465581568
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 152.59866599808447,
"inferenceMs": 140.75054199201986,
"imagesPerSecond": 7.104768378488354,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 152.59866599808447,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1465581568
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 655.7311249925988,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 227.7133750030771,
"inferenceMs": 208.26291700359434,
"imagesPerSecond": 4.8016229407885485,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2209.3813340034103,
"totalMs": 883.4444999956759,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1475706880
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 152.6914999994915,
"inferenceMs": 140.9053330135066,
"imagesPerSecond": 7.096963462015622,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 152.6914999994915,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1475706880
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 153.45095899829175,
"inferenceMs": 141.72008300374728,
"imagesPerSecond": 7.056162957324536,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 153.45095899829175,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1475706880
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 153.48345899838023,
"inferenceMs": 141.70308300526813,
"imagesPerSecond": 7.0570094791997064,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 153.48345899838023,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1475706880
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 154.57020900794305,
"inferenceMs": 142.7658330067061,
"imagesPerSecond": 7.004477044259094,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 154.57020900794305,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1475706880
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01146902161417529,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632150926251,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 155.62787500675768,
"inferenceMs": 144.00345900503453,
"imagesPerSecond": 6.944277636865923,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 155.62787500675768,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1475706880
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 929.4811250001658,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 563.9459579979302,
"inferenceMs": 526.7359999998007,
"imagesPerSecond": 1.898484250175379,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638612992,
"processInitMs": 2680.3774580039317,
"totalMs": 1493.427082998096,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 3.725290298461914e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01287752932330477,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.421337120271651,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.17004198918585,
"inferenceMs": 40.66308299661614,
"imagesPerSecond": 24.592331085255317,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638612992,
"processInitMs": 0,
"totalMs": 55.17004198918585,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 2.0954757928848267e-8,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0011582449940306105,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.085555987611758,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 58.0739159922814,
"inferenceMs": 41.00741699221544,
"imagesPerSecond": 24.385832450501162,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"processInitMs": 0,
"totalMs": 58.0739159922814,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 2.805609256029129e-8,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0018355680687136555,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.940501251149222,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 58.28558398934547,
"inferenceMs": 41.84212499239948,
"imagesPerSecond": 23.89935980024073,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"processInitMs": 0,
"totalMs": 58.28558398934547,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000010579824447631836,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.16716542521498923,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.287423116754459,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.03345799620729,
"inferenceMs": 41.39549999672454,
"imagesPerSecond": 24.157215158148254,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"processInitMs": 0,
"totalMs": 53.03345799620729,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000003427267074584961,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.23488087859004736,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.268470729558459,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.61029198684264,
"inferenceMs": 42.751000000862405,
"imagesPerSecond": 23.391265700915238,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"processInitMs": 0,
"totalMs": 53.61029198684264,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.7578713595867157e-8,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0011362105142325163,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.6075888356297074,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 45.99262500414625,
"inferenceMs": 40.037499988102354,
"imagesPerSecond": 24.976584459498284,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"processInitMs": 0,
"totalMs": 45.99262500414625,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.42937500285916,
"inferenceMs": 41.330000007292256,
"imagesPerSecond": 24.195499632798455,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 55.42937500285916,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.74170900217723,
"inferenceMs": 42.61258299811743,
"imagesPerSecond": 23.4672467530114,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 55.74170900217723,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.61675000353716,
"inferenceMs": 41.206415990018286,
"imagesPerSecond": 24.268065445008293,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 55.61675000353716,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 57.07487500330899,
"inferenceMs": 42.7680829889141,
"imagesPerSecond": 23.38192245509834,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 57.07487500330899,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.245207997155376,
"inferenceMs": 40.28787500283215,
"imagesPerSecond": 24.821363746032823,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 54.245207997155376,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.01412499870639,
"inferenceMs": 40.678292003576644,
"imagesPerSecond": 24.583136379277548,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 55.01412499870639,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.26741699234117,
"inferenceMs": 42.00025000318419,
"imagesPerSecond": 23.809382085206312,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 56.26741699234117,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.66487501002848,
"inferenceMs": 41.87070799525827,
"imagesPerSecond": 23.883044922795357,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 56.66487501002848,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.85475000750739,
"inferenceMs": 40.54720800195355,
"imagesPerSecond": 24.6626105538961,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 54.85475000750739,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.96662500081584,
"inferenceMs": 40.546874995925464,
"imagesPerSecond": 24.662813104597817,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 54.96662500081584,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.38391700631473,
"inferenceMs": 42.33987499901559,
"imagesPerSecond": 23.618397551321305,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 56.38391700631473,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.14804200013168,
"inferenceMs": 42.55862500576768,
"imagesPerSecond": 23.49699972366299,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 56.14804200013168,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.85279098886531,
"inferenceMs": 42.50149999279529,
"imagesPerSecond": 23.52858134817634,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 55.85279098886531,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.62291700334754,
"inferenceMs": 42.25612498703413,
"imagesPerSecond": 23.665208305466724,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 55.62291700334754,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.331500003580004,
"inferenceMs": 43.19170799863059,
"imagesPerSecond": 23.15259216032173,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 56.331500003580004,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.44045799958985,
"inferenceMs": 41.19716699642595,
"imagesPerSecond": 24.273513760952415,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 54.44045799958985,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 57.296708997455426,
"inferenceMs": 43.447040996397845,
"imagesPerSecond": 23.01652718036446,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 57.296708997455426,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.249750006827526,
"inferenceMs": 43.26391700305976,
"imagesPerSecond": 23.113949666861576,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 56.249750006827526,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.46587500430178,
"inferenceMs": 42.4939170043217,
"imagesPerSecond": 23.532779995270815,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 55.46587500430178,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.5475830003852,
"inferenceMs": 40.707417007070035,
"imagesPerSecond": 24.56554784171938,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 54.5475830003852,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.751208997913636,
"inferenceMs": 41.584750011679716,
"imagesPerSecond": 24.047276939722725,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 54.751208997913636,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.29158400511369,
"inferenceMs": 40.55933300696779,
"imagesPerSecond": 24.655237792697612,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 54.29158400511369,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.85804099205416,
"inferenceMs": 40.66454200074077,
"imagesPerSecond": 24.591448736390134,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 53.85804099205416,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.03570800262969,
"inferenceMs": 40.77274999872316,
"imagesPerSecond": 24.526184768781015,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 54.03570800262969,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.460624989587814,
"inferenceMs": 40.70179199334234,
"imagesPerSecond": 24.568942816168185,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 54.460624989587814,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.26441700547002,
"inferenceMs": 40.785707999020815,
"imagesPerSecond": 24.51839257084879,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 54.26441700547002,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.69749999954365,
"inferenceMs": 40.47895799158141,
"imagesPerSecond": 24.70419323066504,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 53.69749999954365,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.56762500398327,
"inferenceMs": 40.502749994629994,
"imagesPerSecond": 24.689681568105467,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 54.56762500398327,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.80362500040792,
"inferenceMs": 41.468374998657964,
"imagesPerSecond": 24.114762153866963,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 55.80362500040792,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.759834001539275,
"inferenceMs": 40.292958001373336,
"imagesPerSecond": 24.81823250519151,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 53.759834001539275,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 953352192
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 921.8036669917637,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 155.5812500009779,
"inferenceMs": 124.47200001042802,
"imagesPerSecond": 8.033935342215294,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638612992,
"processInitMs": 2409.441041003447,
"totalMs": 1077.3849169927416,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 778813440
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.440957996528596,
"inferenceMs": 42.20162499404978,
"imagesPerSecond": 23.69577001219728,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638612992,
"totalMs": 56.440957996528596,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 778813440
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.82679200253915,
"inferenceMs": 42.27762500522658,
"imagesPerSecond": 23.65317351380014,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638612992,
"totalMs": 56.82679200253915,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 778813440
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.40462500741705,
"inferenceMs": 42.19966700475197,
"imagesPerSecond": 23.69686945367112,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638612992,
"totalMs": 56.40462500741705,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 778813440
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.71549999876879,
"inferenceMs": 42.199792005703785,
"imagesPerSecond": 23.696799260641818,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638612992,
"totalMs": 55.71549999876879,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 778813440
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.25654201139696,
"inferenceMs": 43.063457997050136,
"imagesPerSecond": 23.221544355971144,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 56.25654201139696,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 778813440
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 886.8648750067223,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 163.83791698899586,
"inferenceMs": 131.521666990011,
"imagesPerSecond": 7.603309955582827,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638612992,
"processInitMs": 2559.073875003378,
"totalMs": 1050.7027919957181,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 786677760
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 57.89141698915046,
"inferenceMs": 42.86662500817329,
"imagesPerSecond": 23.328171970835868,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638612992,
"totalMs": 57.89141698915046,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 786677760
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 58.23691601108294,
"inferenceMs": 43.29204199893866,
"imagesPerSecond": 23.098933518185994,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638612992,
"totalMs": 58.23691601108294,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 786677760
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 57.09770800604019,
"inferenceMs": 42.44583399849944,
"imagesPerSecond": 23.55943813085054,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638612992,
"totalMs": 57.09770800604019,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 786677760
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.69050000142306,
"inferenceMs": 42.52187500242144,
"imagesPerSecond": 23.517307266978566,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638612992,
"totalMs": 56.69050000142306,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 786677760
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.3085386753082275e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011469074535852997,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.615632954363319,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.88374999933876,
"inferenceMs": 41.64141698856838,
"imagesPerSecond": 24.014552633368005,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1638596608,
"totalMs": 55.88374999933876,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 786677760
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002664271742105484,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011734523377526784,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.617187457331514,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 793.4297909960151,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 608.1138329900568,
"inferenceMs": 577.0029999985127,
"imagesPerSecond": 1.7330932421539882,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"processInitMs": 2311.1286659986945,
"totalMs": 1401.543623986072,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1403813888
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00028739310801029205,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.013163148469175212,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.414062546751873,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.09016700589564,
"inferenceMs": 35.87770799640566,
"imagesPerSecond": 27.872460528977577,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"processInitMs": 0,
"totalMs": 51.09016700589564,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1403813888
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00004735216498374939,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0012055866158107165,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.093749956862302,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.42550000164192,
"inferenceMs": 35.84962499735411,
"imagesPerSecond": 27.894294572782982,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"processInitMs": 0,
"totalMs": 51.42550000164192,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1403813888
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00005954247899353504,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.001895089564754926,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.945312517704877,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.30966599972453,
"inferenceMs": 35.937332999310456,
"imagesPerSecond": 27.826216264272794,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160445952,
"processInitMs": 0,
"totalMs": 51.30966599972453,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1403813888
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0005803853273391724,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.16774411432379566,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.257812487015716,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 47.31066699605435,
"inferenceMs": 35.92387499520555,
"imagesPerSecond": 27.836640677918552,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160445952,
"processInitMs": 0,
"totalMs": 47.31066699605435,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1403813888
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.004442930221557617,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.23931934371466923,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.257812591949104,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 45.58866699517239,
"inferenceMs": 35.06587501033209,
"imagesPerSecond": 28.517754075874393,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160445952,
"processInitMs": 0,
"totalMs": 45.58866699517239,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1403813888
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000022447551600635052,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0011523857538122684,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.6015624469280079,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 39.55237500485964,
"inferenceMs": 34.169499995186925,
"imagesPerSecond": 29.265865761596125,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160445952,
"processInitMs": 0,
"totalMs": 39.55237500485964,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1403813888
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002664271742105484,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011734523377526784,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.617187457331514,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.14470799465198,
"inferenceMs": 34.16937500878703,
"imagesPerSecond": 29.26597281170168,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160445952,
"totalMs": 48.14470799465198,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1403813888
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002664271742105484,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011734523377526784,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.617187457331514,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.31541600287892,
"inferenceMs": 34.26612498878967,
"imagesPerSecond": 29.183340699514606,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160445952,
"totalMs": 48.31541600287892,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1403813888
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002664271742105484,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011734523377526784,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.617187457331514,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 47.78045800048858,
"inferenceMs": 34.09087499312591,
"imagesPerSecond": 29.333362672610786,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160445952,
"totalMs": 47.78045800048858,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1403813888
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002664271742105484,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011734523377526784,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.617187457331514,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.18866700225044,
"inferenceMs": 34.19708399451338,
"imagesPerSecond": 29.242259373940808,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160445952,
"totalMs": 48.18866700225044,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1403813888
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002664271742105484,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011734523377526784,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.617187457331514,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 49.14704199472908,
"inferenceMs": 35.218707998865284,
"imagesPerSecond": 28.39400014424774,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160445952,
"totalMs": 49.14704199472908,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1403813888
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002664271742105484,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011734523377526784,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.617187457331514,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 834.7343339992221,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 137.5364580017049,
"inferenceMs": 104.49275000428315,
"imagesPerSecond": 9.57004194031653,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"processInitMs": 2308.290250002756,
"totalMs": 972.270792000927,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1317584896
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002664271742105484,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011734523377526784,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.617187457331514,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.55554200184997,
"inferenceMs": 33.98374999233056,
"imagesPerSecond": 29.425828527625104,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"totalMs": 48.55554200184997,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1317584896
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002664271742105484,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011734523377526784,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.617187457331514,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.128791007911786,
"inferenceMs": 33.77791699313093,
"imagesPerSecond": 29.605141140093387,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"totalMs": 48.128791007911786,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1317584896
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002664271742105484,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011734523377526784,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.617187457331514,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.96879200532567,
"inferenceMs": 34.407875005854294,
"imagesPerSecond": 29.06311418039783,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"totalMs": 48.96879200532567,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1317584896
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002664271742105484,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011734523377526784,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.617187457331514,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 52.37391700211447,
"inferenceMs": 35.582666998379864,
"imagesPerSecond": 28.10357076510121,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"totalMs": 52.37391700211447,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1317584896
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002664271742105484,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011734523377526784,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.617187457331514,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 49.490916993818246,
"inferenceMs": 34.360917008598335,
"imagesPerSecond": 29.102832143559034,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"totalMs": 49.490916993818246,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1317584896
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002664271742105484,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011734523377526784,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.617187457331514,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 825.2817500033416,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 125.92195799516048,
"inferenceMs": 94.05758300272282,
"imagesPerSecond": 10.631784998887879,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"processInitMs": 2341.0540409968235,
"totalMs": 951.2037079985021,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1367687168
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002664271742105484,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011734523377526784,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.617187457331514,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.574999993434176,
"inferenceMs": 34.475709006073885,
"imagesPerSecond": 29.005929938201454,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"totalMs": 48.574999993434176,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1367687168
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002664271742105484,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011734523377526784,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.617187457331514,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 49.53787500562612,
"inferenceMs": 35.078667002380826,
"imagesPerSecond": 28.50735462473899,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"totalMs": 49.53787500562612,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1367687168
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002664271742105484,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011734523377526784,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.617187457331514,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 47.93341699405573,
"inferenceMs": 33.74508299748413,
"imagesPerSecond": 29.633946968645922,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"totalMs": 47.93341699405573,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1367687168
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002664271742105484,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011734523377526784,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.617187457331514,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 49.00574999919627,
"inferenceMs": 34.69533300085459,
"imagesPerSecond": 28.822320280810356,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"totalMs": 49.00574999919627,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1367687168
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002664271742105484,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.011734523377526784,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.617187457331514,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.01045799104031,
"inferenceMs": 34.16987499804236,
"imagesPerSecond": 29.265544578588344,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"totalMs": 48.01045799104031,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1367687168
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0009202351793646812,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.012386252554279054,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.624999975230081,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 837.937582997256,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 548.0035829968983,
"inferenceMs": 517.5865829951363,
"imagesPerSecond": 1.9320438992318256,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"processInitMs": 2370.219417003682,
"totalMs": 1385.9411659941543,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1221165056
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0011281976476311684,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014001576033479068,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.437499990478973,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.31862499122508,
"inferenceMs": 34.05895900505129,
"imagesPerSecond": 29.360850396269885,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"processInitMs": 0,
"totalMs": 48.31862499122508,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1221165056
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00008560577407479286,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.00124384022490176,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.1249999744424946,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.34241600171663,
"inferenceMs": 35.77420799410902,
"imagesPerSecond": 27.953099623188617,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"processInitMs": 0,
"totalMs": 51.34241600171663,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1221165056
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00008935714140534401,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0019248608189172955,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.937500027417866,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 50.20233399409335,
"inferenceMs": 34.22195800521877,
"imagesPerSecond": 29.221004825249988,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160445952,
"processInitMs": 0,
"totalMs": 50.20233399409335,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1221165056
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006105244159698486,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.17326675137883285,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.187500023969559,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 45.5880000081379,
"inferenceMs": 34.556207989226095,
"imagesPerSecond": 28.93836037541443,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160445952,
"processInitMs": 0,
"totalMs": 45.5880000081379,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1221165056
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.012206286191940308,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.22266893200867344,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.124999982927532,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 46.21029199915938,
"inferenceMs": 35.14908399665728,
"imagesPerSecond": 28.4502435424804,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160445952,
"processInitMs": 0,
"totalMs": 46.21029199915938,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1221165056
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00032257719431072474,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0008375407487619668,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.4374999977941387,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 38.666084001306444,
"inferenceMs": 33.61483300977852,
"imagesPerSecond": 29.74877190998096,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160445952,
"processInitMs": 0,
"totalMs": 38.666084001306444,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1221165056
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0009202351793646812,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.012386252554279054,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.624999975230081,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.241166994557716,
"inferenceMs": 34.295875011594035,
"imagesPerSecond": 29.158025554441775,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160445952,
"totalMs": 48.241166994557716,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1221165056
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0009202351793646812,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.012386252554279054,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.624999975230081,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.64066700974945,
"inferenceMs": 35.045625001657754,
"imagesPerSecond": 28.53423216029668,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160445952,
"totalMs": 48.64066700974945,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1221165056
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0009202351793646812,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.012386252554279054,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.624999975230081,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.74649998964742,
"inferenceMs": 35.353208004380576,
"imagesPerSecond": 28.28597619418558,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160445952,
"totalMs": 48.74649998964742,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1221165056
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0009202351793646812,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.012386252554279054,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.624999975230081,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 47.66237500007264,
"inferenceMs": 33.871958003146574,
"imagesPerSecond": 29.522946382583015,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160445952,
"totalMs": 47.66237500007264,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1221165056
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0009202351793646812,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.012386252554279054,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.624999975230081,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 47.5283749983646,
"inferenceMs": 33.766500011552125,
"imagesPerSecond": 29.61515110117667,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160445952,
"totalMs": 47.5283749983646,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1221165056
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0009202351793646812,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.012386252554279054,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.624999975230081,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 858.805083000334,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 128.4333340008743,
"inferenceMs": 99.80566700687632,
"imagesPerSecond": 10.019471138157945,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"processInitMs": 2426.3316669967026,
"totalMs": 987.2384170012083,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1425555456
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0009202351793646812,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.012386252554279054,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.624999975230081,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 47.79395798686892,
"inferenceMs": 33.792250003898516,
"imagesPerSecond": 29.59258409501092,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"totalMs": 47.79395798686892,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1425555456
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0009202351793646812,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.012386252554279054,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.624999975230081,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 46.99770799197722,
"inferenceMs": 33.55379099957645,
"imagesPerSecond": 29.80289172131468,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"totalMs": 46.99770799197722,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1425555456
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0009202351793646812,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.012386252554279054,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.624999975230081,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.68283298856113,
"inferenceMs": 34.91412500443403,
"imagesPerSecond": 28.64170303202506,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"totalMs": 48.68283298856113,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1425555456
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0009202351793646812,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.012386252554279054,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.624999975230081,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 47.088874998735264,
"inferenceMs": 33.60620800231118,
"imagesPerSecond": 29.756406909438507,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"totalMs": 47.088874998735264,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1425555456
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0009202351793646812,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.012386252554279054,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.624999975230081,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 47.82350000459701,
"inferenceMs": 34.14883300138172,
"imagesPerSecond": 29.28357756645852,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"totalMs": 47.82350000459701,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1425555456
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0009202351793646812,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.012386252554279054,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.624999975230081,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 885.276041997713,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 130.44054199417587,
"inferenceMs": 99.75070800282992,
"imagesPerSecond": 10.024991501530295,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"processInitMs": 2451.830999998492,
"totalMs": 1015.7165839918889,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1395458048
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0009202351793646812,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.012386252554279054,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.624999975230081,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 47.87095800566021,
"inferenceMs": 33.67245799745433,
"imagesPerSecond": 29.69786167899002,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"totalMs": 47.87095800566021,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1395458048
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0009202351793646812,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.012386252554279054,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.624999975230081,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.7133749993518,
"inferenceMs": 33.777165997889824,
"imagesPerSecond": 29.605799375307964,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"totalMs": 48.7133749993518,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1395458048
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0009202351793646812,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.012386252554279054,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.624999975230081,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.956541999359615,
"inferenceMs": 34.453583997674286,
"imagesPerSecond": 29.024556634441943,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"totalMs": 48.956541999359615,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1395458048
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0009202351793646812,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.012386252554279054,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.624999975230081,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.05779100570362,
"inferenceMs": 33.79612498974893,
"imagesPerSecond": 29.589191077477693,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"totalMs": 48.05779100570362,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1395458048
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0009202351793646812,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.012386252554279054,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.624999975230081,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 50.548916013212875,
"inferenceMs": 35.28700000606477,
"imagesPerSecond": 28.339048369884946,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
384,
384
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1160462336,
"totalMs": 50.548916013212875,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "f775b65a79762255128c981547af89addcfe0f88",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1395458048
}
}
]SigLIP 2 Base — Image & Text Matching (512) Desktop
145 configurations testedQuality criteria
{ "minimumTop1Accuracy": 1, "maximumCpuScoreDifference": 0.03, "finiteScoresInUnitInterval": true, "scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification", "performanceTargets": { "goodInferenceMs": 250, "acceptableInferenceMs": 1000, "description": "One512x512 image and four labels; good below250ms, acceptable below1s, slower cases retained but not recommended" }, "version": 2, "minimumLogOddsMargin": 1, "gateRevisionReason": "V1 required0.05absolute sigmoid margin, which conflated calibrated confidence with ranking despite the native independent sigmoid contract. All six semantic fixtures ranked correctly but dog/portrait failed absolute margin. Preserve fullv1failures;v2requires expected top1plus ≥1natural-log-odds separation (odds ratio≥e), finite scores and unchanged0.03CPU agreement. This measures ranking separation without treating sigmoid scores as calibrated probabilities." }
| Precision | Device | Result | Platform | Inference / load | |
|---|---|---|---|---|---|
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 63 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 64 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 64 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 64 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 64 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 64 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 64 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 64 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 64 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 64 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 64 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 64 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 64 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 64 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 64 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 64 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 64 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 64 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 65 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 65 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 65 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 65 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 65 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 65 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 65 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 65 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 66 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 66 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 66 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 66 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 66 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 66 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 66 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 66 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 66 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 67 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 67 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 68 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 68 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 68 ms / 0 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 68 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 68 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 70 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 70 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 72 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 72 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 72 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 73 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 73 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 73 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 73 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 73 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 73 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 73 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 73 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 73 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 73 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 74 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 74 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 74 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 74 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 74 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 75 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 75 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 75 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 75 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 75 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 75 ms / 0 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 76 ms / 0 ms |
BF16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 110 ms / 821 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 111 ms / 820 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 119 ms / 855 ms |
BF16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 120 ms / 802 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 137 ms / 867 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 139 ms / 892 ms |
BF16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 169 ms / 818 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 196 ms / 1043 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 207 ms / 866 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 216 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 217 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 217 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 218 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 218 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 218 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 219 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 219 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 219 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 219 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 220 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 222 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 223 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 223 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 223 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 226 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 226 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 230 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 231 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 234 ms / 0 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 240 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 282 ms / 639 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 285 ms / 678 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 293 ms / 624 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 450 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 450 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 450 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 451 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 453 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 454 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 455 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 455 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 456 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 456 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 456 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 457 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 458 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 458 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 459 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 459 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 460 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 460 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 460 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 462 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 465 ms / 0 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 515 ms / 700 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 523 ms / 633 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 525 ms / 642 ms |
Test setup & sample details
- Quality gate
- { "minimumTop1Accuracy": 1, "maximumCpuScoreDifference": 0.03, "finiteScoresInUnitInterval": true, "scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification", "performanceTargets": { "goodInferenceMs": 250, "acceptableInferenceMs": 1000, "description": "One512x512 image and four labels; good below250ms, acceptable below1s, slower cases retained but not recommended" }, "version": 2, "minimumLogOddsMargin": 1, "gateRevisionReason": "V1 required0.05absolute sigmoid margin, which conflated calibrated confidence with ranking despite the native independent sigmoid contract. All six semantic fixtures ranked correctly but dog/portrait failed absolute margin. Preserve fullv1failures;v2requires expected top1plus ≥1natural-log-odds separation (odds ratio≥e), finite scores and unchanged0.03CPU agreement. This measures ranking separation without treating sigmoid scores as calibrated probabilities." }
- Tested
- 2026-11-02T17:31:21.704594+00:00
- Environment
- { "hardware": "Apple M3 Max", "os": "macOS-27.2-arm64-arm-64bit-Mach-O", "runtime": "Transformers5.22.1/Torch2.10.0/Pillow12.0.0", "interpreterScope": "Existing isolated private probe venv from app-ownedCPython3.13.18; managed packaged provision is separate evidence" }
- Runtime
- transformers 5.22.1
- Artifact revision
- a89f5c5093f902bf39d3cd4d81d2c09867f0724b
- Fixture
{ "file": "model-samples/desktop/dinov2/dinov2-dog.png", "size": 709899, "sha1": "409baef01589c0a55c12bba3b4f5330adc74b1c7", "sha256": "cc986ccf120f7892c9ee0202001c30a91a3039cc5859d851b981476507a25fe3", "url": "https://assets.localai.tools/samples/v1/files/cc986ccf120f7892c9ee0202001c30a91a3039cc5859d851b981476507a25fe3/dinov2-dog.png" }- Rerun when
- Original weights/processor/tokenizer artifact revision; Pinned runtime or relevant OS/driver/backend behavior; Fixture bytes/labels/expected result or quality gate; Image preprocessing/prompt template/sigmoid scoring
Per-case quality & failure details
[
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 641.9438330049161,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 545.9844999859342,
"inferenceMs": 525.0065829895902,
"imagesPerSecond": 1.9047380211989224,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2308.4277500020107,
"totalMs": 1187.9283329908503,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1502396416
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010548985585046466,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.560079563990053,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 470.93979200872127,
"inferenceMs": 457.41849999467377,
"imagesPerSecond": 2.18618180071782,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 470.93979200872127,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1502396416
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0008825274596802046,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.335257469534077,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 479.6381670021219,
"inferenceMs": 464.6624160086503,
"imagesPerSecond": 2.1521000312222016,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 479.6381670021219,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1502396416
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.00136368643836704,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.156745016942408,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 470.80854199884925,
"inferenceMs": 455.6619170034537,
"imagesPerSecond": 2.1946095617914465,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 470.80854199884925,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1502396416
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.10491062300297926,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.758288261772158,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 469.4211250025546,
"inferenceMs": 460.1386249996722,
"imagesPerSecond": 2.1732581132494855,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 469.4211250025546,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1502396416
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.14351956693008106,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.738910594962809,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 467.6557079947088,
"inferenceMs": 458.50833399163093,
"imagesPerSecond": 2.1809854387908527,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 467.6557079947088,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1502396416
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0008568293997086585,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.3198346693234955,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 459.93337500840425,
"inferenceMs": 454.78374999947846,
"imagesPerSecond": 2.198847254329441,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 459.93337500840425,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1502396416
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 472.78204098984133,
"inferenceMs": 458.90591701027006,
"imagesPerSecond": 2.1790958951126806,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 472.78204098984133,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1502396416
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 467.6804170012474,
"inferenceMs": 454.23466598731466,
"imagesPerSecond": 2.2015052458103823,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 467.6804170012474,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1502396416
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 466.54320899688173,
"inferenceMs": 453.13208299921826,
"imagesPerSecond": 2.206862055277876,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 466.54320899688173,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1502396416
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 470.05908298888244,
"inferenceMs": 456.48712500405964,
"imagesPerSecond": 2.1906422880844816,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 470.05908298888244,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1502396416
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 473.5619579878403,
"inferenceMs": 460.01954199164174,
"imagesPerSecond": 2.1738206939438443,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 473.5619579878403,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1502396416
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 699.6843329980038,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 536.0229169891682,
"inferenceMs": 515.1874589937506,
"imagesPerSecond": 1.9410410376703877,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2235.46912499296,
"totalMs": 1235.707249987172,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1490272256
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 468.60845800256357,
"inferenceMs": 454.9022499995772,
"imagesPerSecond": 2.198274464461166,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 468.60845800256357,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1490272256
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 463.95745799236465,
"inferenceMs": 450.60133301012684,
"imagesPerSecond": 2.219256639388428,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 463.95745799236465,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1490272256
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 462.94629099429585,
"inferenceMs": 450.09741600370035,
"imagesPerSecond": 2.2217412596560626,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 462.94629099429585,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1490272256
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 462.91383300558664,
"inferenceMs": 449.6590000053402,
"imagesPerSecond": 2.223907449841155,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 462.91383300558664,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1490272256
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 462.895832999493,
"inferenceMs": 449.63387500320096,
"imagesPerSecond": 2.2240317191245453,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 462.895832999493,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1490272256
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 633.2416249933885,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 545.9124160115607,
"inferenceMs": 522.717083003954,
"imagesPerSecond": 1.9130807706784583,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2212.343207997037,
"totalMs": 1179.1540410049492,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1488715776
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 475.6380829931004,
"inferenceMs": 462.0655829930911,
"imagesPerSecond": 2.1641949472245203,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 475.6380829931004,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1488715776
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 472.7452499937499,
"inferenceMs": 459.5008750038687,
"imagesPerSecond": 2.1762744194808783,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 472.7452499937499,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1488715776
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 471.2028750072932,
"inferenceMs": 458.07070899172686,
"imagesPerSecond": 2.183069077263486,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 471.2028750072932,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1488715776
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 471.4725830126554,
"inferenceMs": 458.08629199746065,
"imagesPerSecond": 2.182994814447631,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 471.4725830126554,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1488715776
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 470.58241700869985,
"inferenceMs": 456.49662498908583,
"imagesPerSecond": 2.190596699425562,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 470.58241700869985,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1488715776
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 639.4549580145394,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 304.05345899634995,
"inferenceMs": 282.13812499598134,
"imagesPerSecond": 3.5443632441175525,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2126.684499991825,
"totalMs": 943.5084170108894,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1523351552
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010548985585046466,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.560079563990053,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 231.6712500032736,
"inferenceMs": 218.2939580088714,
"imagesPerSecond": 4.580978828371238,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 231.6712500032736,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1523351552
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0008825274596802046,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.335257469534077,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 233.12366699974518,
"inferenceMs": 217.8213750012219,
"imagesPerSecond": 4.590917672769215,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 233.12366699974518,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1523351552
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.00136368643836704,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.156745016942408,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 231.8328750116052,
"inferenceMs": 216.8114579981193,
"imagesPerSecond": 4.6123023627684585,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 231.8328750116052,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1523351552
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.10491062300297926,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.758288261772158,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 227.34183399006724,
"inferenceMs": 217.93124999385327,
"imagesPerSecond": 4.588603057286208,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 227.34183399006724,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1523351552
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.14351956693008106,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.738910594962809,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 227.6883330050623,
"inferenceMs": 218.51145799155347,
"imagesPerSecond": 4.576419054595548,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 227.6883330050623,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1523351552
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0008568293997086585,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.3198346693234955,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 221.5465830086032,
"inferenceMs": 216.69954199751373,
"imagesPerSecond": 4.614684418721445,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 0,
"totalMs": 221.5465830086032,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1523351552
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 236.4715420117136,
"inferenceMs": 223.3880840067286,
"imagesPerSecond": 4.4765145126088255,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 236.4715420117136,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1523351552
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 247.3163749964442,
"inferenceMs": 234.1454590059584,
"imagesPerSecond": 4.270849429433319,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 247.3163749964442,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1523351552
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 254.25250000262167,
"inferenceMs": 239.8381669918308,
"imagesPerSecond": 4.169478163306932,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 254.25250000262167,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1523351552
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 229.71695799787994,
"inferenceMs": 216.2852500041481,
"imagesPerSecond": 4.623523795454481,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 229.71695799787994,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1523351552
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 233.49599998618942,
"inferenceMs": 220.09870799956843,
"imagesPerSecond": 4.543416038598286,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 233.49599998618942,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1523351552
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 677.6521660067374,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 305.5618750076974,
"inferenceMs": 284.70025000569876,
"imagesPerSecond": 3.512466181466238,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2326.1403329961468,
"totalMs": 983.2140410144348,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1502445568
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 232.10725000535604,
"inferenceMs": 218.5556249896763,
"imagesPerSecond": 4.575494225084512,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 232.10725000535604,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1502445568
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 239.08483400009573,
"inferenceMs": 225.60783399967477,
"imagesPerSecond": 4.432470195168142,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 239.08483400009573,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1502445568
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 236.1665000062203,
"inferenceMs": 222.7134999993723,
"imagesPerSecond": 4.490073569868097,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 236.1665000062203,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1502445568
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 236.0711670044111,
"inferenceMs": 222.40195899212267,
"imagesPerSecond": 4.496363271851483,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 236.0711670044111,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1502445568
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 245.07416700362228,
"inferenceMs": 231.20075000042561,
"imagesPerSecond": 4.3252454847060795,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 245.07416700362228,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1502445568
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 624.4705419958336,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 313.5266669996781,
"inferenceMs": 292.6204579998739,
"imagesPerSecond": 3.417396059165593,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2081.33558400732,
"totalMs": 937.9972089955118,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1506230272
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 239.41924999235198,
"inferenceMs": 225.81524999986868,
"imagesPerSecond": 4.428398879174819,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 239.41924999235198,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1506230272
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 231.8772500002524,
"inferenceMs": 218.54187500139233,
"imagesPerSecond": 4.575782101227186,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 231.8772500002524,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1506230272
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 236.79795900534373,
"inferenceMs": 223.41187500569504,
"imagesPerSecond": 4.476037811215312,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 236.79795900534373,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1506230272
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 232.40237499703653,
"inferenceMs": 219.12658400833607,
"imagesPerSecond": 4.563572259046204,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 232.40237499703653,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1506230272
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830769187828992,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789990439924862,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 243.02458300371654,
"inferenceMs": 229.7725829994306,
"imagesPerSecond": 4.3521293400025804,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 243.02458300371654,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1506230272
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 1043.4778339986224,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 233.33158300374635,
"inferenceMs": 196.3714170124149,
"imagesPerSecond": 5.092390813357417,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814738432,
"processInitMs": 2552.199790996383,
"totalMs": 1276.8094170023687,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010548975107667502,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.560072933524354,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 93.78508300869726,
"inferenceMs": 76.03704099892639,
"imagesPerSecond": 13.151484945529635,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814738432,
"processInitMs": 0,
"totalMs": 93.78508300869726,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 6.752088665962219e-9,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0008825342112572798,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.33526423458009,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 89.38879199558869,
"inferenceMs": 71.12245900498237,
"imagesPerSecond": 14.060256267713502,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814738432,
"processInitMs": 0,
"totalMs": 89.38879199558869,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 6.51925802230835e-9,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0013636799282039647,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.156748770240943,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 92.27970799838658,
"inferenceMs": 73.67291700211354,
"imagesPerSecond": 13.57350897306417,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"processInitMs": 0,
"totalMs": 92.27970799838658,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 9.685754776000977e-8,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.10491071973137878,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.758270235192931,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 87.5778750050813,
"inferenceMs": 75.36658299795818,
"imagesPerSecond": 13.268480010923302,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"processInitMs": 0,
"totalMs": 87.5778750050813,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000016391277313232422,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.14351792772686167,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.73888961885007,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 81.82179099821951,
"inferenceMs": 70.50025000353344,
"imagesPerSecond": 14.1843468633073,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"processInitMs": 0,
"totalMs": 81.82179099821951,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 7.799826562404633e-9,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0008568257908336818,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.3198413968959715,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 78.524374999688,
"inferenceMs": 71.71112499781884,
"imagesPerSecond": 13.944837708659792,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"processInitMs": 0,
"totalMs": 78.524374999688,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 90.7062079932075,
"inferenceMs": 73.52516699756961,
"imagesPerSecond": 13.600785157455748,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 90.7062079932075,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 86.43070899415761,
"inferenceMs": 70.5026670038933,
"imagesPerSecond": 14.183860589909004,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 86.43070899415761,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 89.81525000126567,
"inferenceMs": 74.01412499893922,
"imagesPerSecond": 13.510934568426393,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 89.81525000126567,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 87.13670799625106,
"inferenceMs": 70.82620800065342,
"imagesPerSecond": 14.119067337203402,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 87.13670799625106,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 87.66829199157655,
"inferenceMs": 71.30320799478795,
"imagesPerSecond": 14.024614433520256,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 87.66829199157655,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 89.4855839869706,
"inferenceMs": 72.8615410043858,
"imagesPerSecond": 13.724661683175304,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 89.4855839869706,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 89.05337499163579,
"inferenceMs": 72.50062499952037,
"imagesPerSecond": 13.792984543327945,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 89.05337499163579,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 90.47749999444932,
"inferenceMs": 74.21533300657757,
"imagesPerSecond": 13.474304560640748,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 90.47749999444932,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 87.34779099177103,
"inferenceMs": 70.98150000092573,
"imagesPerSecond": 14.088177905326853,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 87.34779099177103,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 88.41499999107327,
"inferenceMs": 72.16512500599492,
"imagesPerSecond": 13.857108955564446,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 88.41499999107327,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 91.62649999780115,
"inferenceMs": 74.78637500025798,
"imagesPerSecond": 13.371419593429291,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 91.62649999780115,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 87.77666700189002,
"inferenceMs": 71.10187500074971,
"imagesPerSecond": 14.064326714161275,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 87.77666700189002,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 89.19183300167788,
"inferenceMs": 73.17975000478327,
"imagesPerSecond": 13.664982456685582,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 89.19183300167788,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 88.94162499927916,
"inferenceMs": 72.55883299512789,
"imagesPerSecond": 13.781919564047385,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 88.94162499927916,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 88.73816700361203,
"inferenceMs": 72.65754099353217,
"imagesPerSecond": 13.76319630868072,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 88.73816700361203,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 89.69316699949559,
"inferenceMs": 73.10062499891501,
"imagesPerSecond": 13.679773599949966,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 89.69316699949559,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 87.25720799702685,
"inferenceMs": 71.04750000871718,
"imagesPerSecond": 14.075090606668848,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 87.25720799702685,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 88.86979099770542,
"inferenceMs": 72.50066699634772,
"imagesPerSecond": 13.792976553586408,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 88.86979099770542,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 91.51254101016093,
"inferenceMs": 74.95974999619648,
"imagesPerSecond": 13.340492731775932,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 91.51254101016093,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 91.00445899821352,
"inferenceMs": 75.28566700057127,
"imagesPerSecond": 13.28274079038726,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 91.00445899821352,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 92.12745798868127,
"inferenceMs": 75.23566599411424,
"imagesPerSecond": 13.291568390957435,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 92.12745798868127,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 91.30741700937506,
"inferenceMs": 75.4559170018183,
"imagesPerSecond": 13.25277115081515,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 91.30741700937506,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 89.64099999866448,
"inferenceMs": 73.7555829982739,
"imagesPerSecond": 13.5582956482549,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 89.64099999866448,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 88.72279099887237,
"inferenceMs": 73.10749999305699,
"imagesPerSecond": 13.678487160619223,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 88.72279099887237,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 86.0850000026403,
"inferenceMs": 70.7042919966625,
"imagesPerSecond": 14.1434129634903,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 86.0850000026403,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 88.51325001160149,
"inferenceMs": 72.8733750001993,
"imagesPerSecond": 13.722432918706799,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 88.51325001160149,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 88.60424999147654,
"inferenceMs": 72.9324580024695,
"imagesPerSecond": 13.711316297143583,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 88.60424999147654,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 86.24975000566337,
"inferenceMs": 70.44804199540522,
"imagesPerSecond": 14.194858674215846,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 86.24975000566337,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 86.62162499967963,
"inferenceMs": 71.03166700107977,
"imagesPerSecond": 14.078227954086996,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 86.62162499967963,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 86.1095840082271,
"inferenceMs": 70.81629200547468,
"imagesPerSecond": 14.121044348420442,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 86.1095840082271,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962510848
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 866.8092919979244,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 167.79974999371916,
"inferenceMs": 136.51695899898186,
"imagesPerSecond": 7.3250972431011885,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814738432,
"processInitMs": 2422.1419579989742,
"totalMs": 1034.6090419916436,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 800030720
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 87.26041699992493,
"inferenceMs": 71.13579200813547,
"imagesPerSecond": 14.057620949600656,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814738432,
"totalMs": 87.26041699992493,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 800030720
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 87.73566699528601,
"inferenceMs": 70.82129101036116,
"imagesPerSecond": 14.120047597744298,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814738432,
"totalMs": 87.73566699528601,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 800030720
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 87.61812500597443,
"inferenceMs": 71.66737499937881,
"imagesPerSecond": 13.953350461191969,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814738432,
"totalMs": 87.61812500597443,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 800030720
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 87.1939999924507,
"inferenceMs": 71.49970800674055,
"imagesPerSecond": 13.986071102636199,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 87.1939999924507,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 800030720
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 86.4134580042446,
"inferenceMs": 70.4063749872148,
"imagesPerSecond": 14.203259295505434,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 86.4134580042446,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 800030720
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 892.4856250087032,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 169.9715420108987,
"inferenceMs": 138.50791699951515,
"imagesPerSecond": 7.219803904808565,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814738432,
"processInitMs": 2588.676084007602,
"totalMs": 1062.457167019602,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 796213248
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 86.24808400054462,
"inferenceMs": 70.68550000258256,
"imagesPerSecond": 14.147173040630173,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814738432,
"totalMs": 86.24808400054462,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 796213248
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 85.95975000935141,
"inferenceMs": 70.87374999537133,
"imagesPerSecond": 14.109596290097656,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814738432,
"totalMs": 85.95975000935141,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 796213248
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 86.26866599661298,
"inferenceMs": 70.9118750091875,
"imagesPerSecond": 14.102010416033108,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814738432,
"totalMs": 86.26866599661298,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 796213248
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 86.5545840060804,
"inferenceMs": 70.92270799330436,
"imagesPerSecond": 14.09985642531314,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 86.5545840060804,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 796213248
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.0244548320770264e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009830758830503328,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789985687536011,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 86.63729199906811,
"inferenceMs": 70.704083002056,
"imagesPerSecond": 14.143454770086207,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 2814722048,
"totalMs": 86.63729199906811,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 796213248
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000013690441846847534,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009829376145717106,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789062526669443,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 865.7977910042973,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 242.962082993472,
"inferenceMs": 207.25400000810623,
"imagesPerSecond": 4.824997346062742,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"processInitMs": 2338.1036659993697,
"totalMs": 1108.7598739977693,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1251459072
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000019688159227371216,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010529152539675124,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.554687401520029,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 83.375583999441,
"inferenceMs": 67.32858299801592,
"imagesPerSecond": 14.852532987802055,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"processInitMs": 0,
"totalMs": 83.375583999441,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1251459072
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.809124559164047e-8,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0008824697981708596,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.335937442901305,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 85.79316698887851,
"inferenceMs": 67.65691599866841,
"imagesPerSecond": 14.780454965161011,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"processInitMs": 0,
"totalMs": 85.79316698887851,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1251459072
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000018891412764787674,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0013447927198058096,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.14062509729586,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 84.1178749979008,
"inferenceMs": 66.85879199358169,
"imagesPerSecond": 14.956896021932284,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"processInitMs": 0,
"totalMs": 84.1178749979008,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1251459072
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00044008344411849976,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.10447046182480335,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.742187554626902,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 79.23666699207388,
"inferenceMs": 66.11537499702536,
"imagesPerSecond": 15.125074917067199,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"processInitMs": 0,
"totalMs": 79.23666699207388,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1251459072
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00034174323081970215,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.1431777287607474,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.726562518476717,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 78.31883399921935,
"inferenceMs": 67.69962501130067,
"imagesPerSecond": 14.771130561403794,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254440960,
"processInitMs": 0,
"totalMs": 78.31883399921935,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1251459072
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000002601766027510166,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0008593543316237628,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.3281250839976195,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 72.99958300427534,
"inferenceMs": 66.03004199860152,
"imagesPerSecond": 15.144621595442564,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254440960,
"processInitMs": 0,
"totalMs": 72.99958300427534,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1251459072
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000013690441846847534,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009829376145717106,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789062526669443,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 82.71441599936225,
"inferenceMs": 67.50966601248365,
"imagesPerSecond": 14.812693634346873,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254440960,
"totalMs": 82.71441599936225,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1251459072
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000013690441846847534,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009829376145717106,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789062526669443,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 81.17316699645016,
"inferenceMs": 66.04658300057054,
"imagesPerSecond": 15.140828708600438,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254440960,
"totalMs": 81.17316699645016,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1251459072
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000013690441846847534,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009829376145717106,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789062526669443,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 81.17449999554083,
"inferenceMs": 65.35037500725593,
"imagesPerSecond": 15.302131011321185,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254440960,
"totalMs": 81.17449999554083,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1251459072
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000013690441846847534,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009829376145717106,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789062526669443,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 79.15383299405221,
"inferenceMs": 63.669333001598716,
"imagesPerSecond": 15.706148515406786,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254440960,
"totalMs": 79.15383299405221,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1251459072
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000013690441846847534,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009829376145717106,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789062526669443,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 79.5623330050148,
"inferenceMs": 63.77804199291859,
"imagesPerSecond": 15.679377553030433,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254440960,
"totalMs": 79.5623330050148,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1251459072
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000013690441846847534,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009829376145717106,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789062526669443,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 855.4797499964479,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 153.0925830011256,
"inferenceMs": 118.95862501114607,
"imagesPerSecond": 8.406284116904537,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"processInitMs": 2374.2750409874134,
"totalMs": 1008.5723329975735,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1156104192
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000013690441846847534,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009829376145717106,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789062526669443,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 81.6026250104187,
"inferenceMs": 65.6111249991227,
"imagesPerSecond": 15.241317688324523,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"totalMs": 81.6026250104187,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1156104192
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000013690441846847534,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009829376145717106,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789062526669443,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 79.95883400144521,
"inferenceMs": 64.29904200194869,
"imagesPerSecond": 15.552331245770246,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"totalMs": 79.95883400144521,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1156104192
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000013690441846847534,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009829376145717106,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789062526669443,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 80.37699999113102,
"inferenceMs": 63.83904200629331,
"imagesPerSecond": 15.664395463538114,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"totalMs": 80.37699999113102,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1156104192
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000013690441846847534,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009829376145717106,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789062526669443,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 81.4072499924805,
"inferenceMs": 65.11962501099333,
"imagesPerSecond": 15.356353784764309,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"totalMs": 81.4072499924805,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1156104192
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000013690441846847534,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009829376145717106,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789062526669443,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 81.10395800031256,
"inferenceMs": 65.0286669988418,
"imagesPerSecond": 15.377833287245618,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254440960,
"totalMs": 81.10395800031256,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1156104192
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000013690441846847534,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009829376145717106,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789062526669443,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 819.824874997721,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 143.02825000777375,
"inferenceMs": 110.52641700371169,
"imagesPerSecond": 9.047610762288786,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"processInitMs": 2439.684249999118,
"totalMs": 962.8531250054948,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1387233280
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000013690441846847534,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009829376145717106,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789062526669443,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 80.84754100127611,
"inferenceMs": 64.7136670013424,
"imagesPerSecond": 15.452686369623533,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"totalMs": 80.84754100127611,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1387233280
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000013690441846847534,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009829376145717106,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789062526669443,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 82.04887500323821,
"inferenceMs": 65.89541600260418,
"imagesPerSecond": 15.175562439130518,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"totalMs": 82.04887500323821,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1387233280
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000013690441846847534,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009829376145717106,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789062526669443,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 83.73045800544787,
"inferenceMs": 67.7985840011388,
"imagesPerSecond": 14.749570580754357,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"totalMs": 83.73045800544787,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1387233280
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000013690441846847534,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009829376145717106,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789062526669443,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 80.15312500356231,
"inferenceMs": 63.724749998073094,
"imagesPerSecond": 15.692489967088736,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"totalMs": 80.15312500356231,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1387233280
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000013690441846847534,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.009829376145717106,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.789062526669443,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 81.44720799464267,
"inferenceMs": 65.633499994874,
"imagesPerSecond": 15.236121798747597,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254440960,
"totalMs": 81.44720799464267,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1387233280
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0004669278860092163,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010294938023434952,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.750000012933725,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 817.6333750016056,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 200.48854099877644,
"inferenceMs": 168.69095800211653,
"imagesPerSecond": 5.92800000571135,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"processInitMs": 2324.436540991883,
"totalMs": 1018.121916000382,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1287847936
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0003967657685279846,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01094154399834224,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.4999999399982915,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 81.09529200010002,
"inferenceMs": 64.2651670059422,
"imagesPerSecond": 15.56052907957955,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"processInitMs": 0,
"totalMs": 81.09529200010002,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1287847936
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000027947186026722193,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0009104034634219715,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.249999865423135,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 82.20479199371766,
"inferenceMs": 64.29604200820904,
"imagesPerSecond": 15.553056903134477,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"processInitMs": 0,
"totalMs": 82.20479199371766,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1287847936
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00004560663364827633,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0014092219053054578,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.124999992864859,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 81.53783399029635,
"inferenceMs": 64.11587499314919,
"imagesPerSecond": 15.596761334175826,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"processInitMs": 0,
"totalMs": 81.53783399029635,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1287847936
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0017731860280036926,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.10668317433373886,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.687499961162604,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 75.80479100579396,
"inferenceMs": 64.2312919953838,
"imagesPerSecond": 15.568735563841194,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"processInitMs": 0,
"totalMs": 75.80479100579396,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1287847936
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.004517748951911926,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.14803641057369532,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.687500066235543,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 75.47137500660028,
"inferenceMs": 63.990082999225706,
"imagesPerSecond": 15.62742151798897,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"processInitMs": 0,
"totalMs": 75.47137500660028,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1287847936
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000012453238014131784,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0008544716110918671,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.3125000413918624,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 70.2657499932684,
"inferenceMs": 63.42949999088887,
"imagesPerSecond": 15.765534966279757,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254440960,
"processInitMs": 0,
"totalMs": 70.2657499932684,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1287847936
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0004669278860092163,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010294938023434952,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.750000012933725,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 81.42104098806158,
"inferenceMs": 65.64070800959598,
"imagesPerSecond": 15.234448718222394,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254440960,
"totalMs": 81.42104098806158,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1287847936
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0004669278860092163,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010294938023434952,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.750000012933725,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 81.09429200703744,
"inferenceMs": 65.08641700202134,
"imagesPerSecond": 15.364188813296387,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254440960,
"totalMs": 81.09429200703744,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1287847936
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0004669278860092163,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010294938023434952,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.750000012933725,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 81.28612500149757,
"inferenceMs": 65.35141699714586,
"imagesPerSecond": 15.301887027846293,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254440960,
"totalMs": 81.28612500149757,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1287847936
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0004669278860092163,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010294938023434952,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.750000012933725,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 79.026082996279,
"inferenceMs": 63.51958400045987,
"imagesPerSecond": 15.743176151669383,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254440960,
"totalMs": 79.026082996279,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1287847936
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0004669278860092163,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010294938023434952,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.750000012933725,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 80.6353750085691,
"inferenceMs": 64.19129099231213,
"imagesPerSecond": 15.578437269936897,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254440960,
"totalMs": 80.6353750085691,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1287847936
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0004669278860092163,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010294938023434952,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.750000012933725,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 821.0858749953331,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 144.11220799956936,
"inferenceMs": 110.38208298850805,
"imagesPerSecond": 9.059441287261363,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"processInitMs": 2394.2514170048526,
"totalMs": 965.1980829949025,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1248935936
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0004669278860092163,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010294938023434952,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.750000012933725,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 81.15095800894778,
"inferenceMs": 65.50379200780299,
"imagesPerSecond": 15.26629175729059,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"totalMs": 81.15095800894778,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1248935936
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0004669278860092163,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010294938023434952,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.750000012933725,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 79.86716600134969,
"inferenceMs": 64.03525000496302,
"imagesPerSecond": 15.616398779148914,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"totalMs": 79.86716600134969,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1248935936
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0004669278860092163,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010294938023434952,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.750000012933725,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 81.36837500205729,
"inferenceMs": 65.60729199554771,
"imagesPerSecond": 15.242208138507877,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"totalMs": 81.36837500205729,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1248935936
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0004669278860092163,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010294938023434952,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.750000012933725,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 79.17945898952894,
"inferenceMs": 63.50620799639728,
"imagesPerSecond": 15.746492060378259,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"totalMs": 79.17945898952894,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1248935936
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0004669278860092163,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010294938023434952,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.750000012933725,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 80.55520900234114,
"inferenceMs": 64.85804200929124,
"imagesPerSecond": 15.418288449977336,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254440960,
"totalMs": 80.55520900234114,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1248935936
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0004669278860092163,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010294938023434952,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.750000012933725,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 802.2852909925859,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 152.62508300656918,
"inferenceMs": 120.11083299876191,
"imagesPerSecond": 8.325643699517993,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"processInitMs": 2249.3989169888664,
"totalMs": 954.9103739991551,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1504100352
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0004669278860092163,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010294938023434952,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.750000012933725,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 80.70229200529866,
"inferenceMs": 64.43487500655465,
"imagesPerSecond": 15.519545896508294,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"totalMs": 80.70229200529866,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1504100352
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0004669278860092163,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010294938023434952,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.750000012933725,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 81.49404200958088,
"inferenceMs": 65.3652500041062,
"imagesPerSecond": 15.298648745888384,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"totalMs": 81.49404200958088,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1504100352
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0004669278860092163,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010294938023434952,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.750000012933725,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 80.00041599734686,
"inferenceMs": 63.841333001619205,
"imagesPerSecond": 15.663833334661685,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"totalMs": 80.00041599734686,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1504100352
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0004669278860092163,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010294938023434952,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.750000012933725,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 79.63116699829698,
"inferenceMs": 63.79945800290443,
"imagesPerSecond": 15.674114346778238,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254457344,
"totalMs": 79.63116699829698,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1504100352
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0004669278860092163,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.010294938023434952,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.750000012933725,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"loadMs": 0,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 83.74145800189581,
"inferenceMs": 68.20979199255817,
"imagesPerSecond": 14.660651657009922,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
512,
512
],
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1254440960,
"totalMs": 83.74145800189581,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "a89f5c5093f902bf39d3cd4d81d2c09867f0724b",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1504100352
}
}
]SigLIP 2 Base · NaFlex — Image & Text Matching Desktop
194 configurations testedQuality criteria
{ "minimumTop1Accuracy": 1, "maximumCpuScoreDifference": 0.03, "finiteScoresInUnitInterval": true, "scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification", "performanceTargets": { "goodInferenceMs": 250, "acceptableInferenceMs": 1000, "description": "One image at pinned256px FixRes or256patch NaFlex default and four labels; good<=250ms,acceptable<=1000ms" }, "version": 2, "minimumLogOddsMargin": 1, "gateRevisionReason": "V1 required0.05absolute sigmoid margin, which conflated calibrated confidence with ranking despite the native independent sigmoid contract. All six semantic fixtures ranked correctly but dog/portrait failed absolute margin. Preserve fullv1failures;v2requires expected top1plus ≥1natural-log-odds separation (odds ratio≥e), finite scores and unchanged0.03CPU agreement. This measures ranking separation without treating sigmoid scores as calibrated probabilities." }
| Precision | Device | Result | Platform | Inference / load | |
|---|---|---|---|---|---|
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 22 ms / 2503 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 2503 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 2448 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 2442 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 2442 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 2464 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 2464 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 2464 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 24 ms / 2448 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 24 ms / 2442 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 24 ms / 2503 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 25 ms / 2448 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 28 ms / 2531 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 29 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 29 ms / 2559 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 29 ms / 2531 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 29 ms / 2442 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 29 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 29 ms / 3474 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 30 ms / 2448 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 30 ms / 2464 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 30 ms / 2559 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 30 ms / 2531 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 30 ms / 2559 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 31 ms / 2442 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 31 ms / 2448 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 31 ms / 2464 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 32 ms / 2448 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 32 ms / 2464 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 33 ms / 3474 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 33 ms / 2503 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 3474 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 2448 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 2442 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 2559 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 2503 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 2503 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 2442 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 2464 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 2448 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 2585 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 2442 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 2585 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 2464 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 2464 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 2503 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 2585 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 37 ms / 2503 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 37 ms / 2503 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 37 ms / 2559 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 37 ms / 2559 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 37 ms / 2503 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 38 ms / 2531 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 38 ms / 2503 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 38 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 38 ms / 3474 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 39 ms / 2448 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 39 ms / 2559 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 39 ms / 2585 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 39 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 39 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 39 ms / 2531 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 2464 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 2585 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 2464 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 2442 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 2585 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 2448 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 3474 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 2503 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 2464 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 2559 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 2531 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 2448 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 2531 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 2442 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 2531 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 2531 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 3474 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 2503 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 2531 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 2585 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 2464 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 3474 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 2442 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 2531 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 3474 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 2585 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 2448 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 44 ms / 2448 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 44 ms / 2559 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 44 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 44 ms / 2559 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 44 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 44 ms / 2559 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 44 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 45 ms / 2531 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 46 ms / 2559 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 46 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 46 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 46 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 47 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 47 ms / 2531 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 48 ms / 3474 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 49 ms / 2585 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 51 ms / 2442 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 52 ms / 2559 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 52 ms / 2585 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 54 ms / 3474 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 56 ms / 2585 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 59 ms / 2585 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 70 ms / 2442 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 78 ms / 2666 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 80 ms / 2666 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 81 ms / 3474 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 84 ms / 2666 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 85 ms / 2666 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 88 ms / 2666 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 92 ms / 2666 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 94 ms / 2666 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 98 ms / 2666 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 100 ms / 2666 ms |
BF16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 100 ms / 2585 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 101 ms / 2666 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 105 ms / 2666 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 107 ms / 2666 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 107 ms / 2666 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 108 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 109 ms / 3474 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 154 ms / 2559 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 163 ms / 2531 ms |
BF16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 165 ms / 2442 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 171 ms / 2503 ms |
BF16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 183 ms / 2585 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 199 ms / 2666 ms |
FP16 | mpsAccelerated | Passed | Moderate | macos · aarch64 | 461 ms / 2448 ms |
BF16 | mpsAccelerated | Passed | Moderate | macos · aarch64 | 463 ms / 2464 ms |
FP32 | mpsAccelerated | Passed | Moderate | macos · aarch64 | 526 ms / 3474 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 84 ms / 2385 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 85 ms / 2385 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 85 ms / 2312 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 85 ms / 2312 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 85 ms / 2312 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 85 ms / 2312 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 85 ms / 2385 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 86 ms / 2353 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 86 ms / 2385 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 86 ms / 2385 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 86 ms / 2385 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 86 ms / 2312 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 86 ms / 2353 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 87 ms / 2385 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 87 ms / 2385 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 87 ms / 2312 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 87 ms / 2312 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 87 ms / 2353 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 87 ms / 2385 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 88 ms / 2353 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 88 ms / 2353 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 88 ms / 2312 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 88 ms / 2385 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 88 ms / 2312 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 89 ms / 2385 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 90 ms / 2312 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 90 ms / 2353 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 90 ms / 2353 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 90 ms / 2312 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 90 ms / 2353 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 91 ms / 2353 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 91 ms / 2353 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 92 ms / 2353 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 92 ms / 2312 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 92 ms / 2353 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 92 ms / 2385 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 92 ms / 2353 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 97 ms / 2385 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 107 ms / 2312 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 146 ms / 2312 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 157 ms / 2385 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 181 ms / 2353 ms |
FP32 | cpu | Failed | — | macos · aarch64 | 1037 ms / 2403 ms |
Test setup & sample details
- Quality gate
- { "minimumTop1Accuracy": 1, "maximumCpuScoreDifference": 0.03, "finiteScoresInUnitInterval": true, "scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification", "performanceTargets": { "goodInferenceMs": 250, "acceptableInferenceMs": 1000, "description": "One image at pinned256px FixRes or256patch NaFlex default and four labels; good<=250ms,acceptable<=1000ms" }, "version": 2, "minimumLogOddsMargin": 1, "gateRevisionReason": "V1 required0.05absolute sigmoid margin, which conflated calibrated confidence with ranking despite the native independent sigmoid contract. All six semantic fixtures ranked correctly but dog/portrait failed absolute margin. Preserve fullv1failures;v2requires expected top1plus ≥1natural-log-odds separation (odds ratio≥e), finite scores and unchanged0.03CPU agreement. This measures ranking separation without treating sigmoid scores as calibrated probabilities." }
- Tested
- 2026-11-06T17:22:26.564237+00:00
- Environment
- { "hardware": "Apple M3 Max", "os": "macOS-27.2-arm64-arm-64bit-Mach-O", "runtime": "Transformers5.22.1/Torch2.10.0/Pillow12.0.0", "interpreterScope": "Existing isolated private probe venv from app-ownedCPython3.13.18; managed packaged provision is separate evidence" }
- Runtime
- transformers 5.22.1
- Artifact revision
- b53b807d3a2d5e2b3911292f2d69e5341cdc064c
- Fixture
{ "file": "model-samples/desktop/siglip2-family/siglip2-base-patch16-naflex-dog-input.json", "size": 567, "sha1": "f9b27f8c462739815e725c2824899b4eb249c8be", "sha256": "d280bd6cb58649d1f878f8ec5d188b9ff0b5feb0fd2127e374996c7fa9f3b7c8", "url": "https://assets.localai.tools/samples/v1/files/d280bd6cb58649d1f878f8ec5d188b9ff0b5feb0fd2127e374996c7fa9f3b7c8/siglip2-base-patch16-naflex-dog-input.json" }- Rerun when
- Original weights/processor/tokenizer artifact revision; Pinned runtime or relevant OS/driver/backend behavior; Fixture bytes/labels/expected result or quality gate; Image preprocessing/prompt template/sigmoid scoring
Per-case quality & failure details
[
{
"precision": "fp32",
"device": "cpu",
"status": "fail",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07378945255550207,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.4808044991854,
"nativeAspectRatioGridAndMask": true,
"passed": false
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2402.875291998498,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 1056.2279999976454,
"inferenceMs": 1036.6392920004728,
"imagesPerSecond": 0.9646556981939518,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2402.875291998498,
"totalMs": 3459.1032919961435,
"initializationReused": false,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "slow",
"observedResidentBytesAtCaseEnd": 1452507136
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07378945255550207,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.4808044991854,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2353.2880830025533,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 201.29841699963436,
"inferenceMs": 181.45066699798917,
"imagesPerSecond": 5.511139840621705,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2353.2880830025533,
"totalMs": 2554.5865000021877,
"initializationReused": false,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454424064
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07365939557166712,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.984287222929037,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2353.2880830025533,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 102.08712500025285,
"inferenceMs": 90.73716700004297,
"imagesPerSecond": 11.020842209009308,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2353.2880830025533,
"totalMs": 2455.375208002806,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454424064
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.00565307897772982,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.457427054963535,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2353.2880830025533,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 105.35500000332831,
"inferenceMs": 91.9469999971625,
"imagesPerSecond": 10.87583064190088,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2353.2880830025533,
"totalMs": 2458.6430830058816,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454424064
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.011110889799056167,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.53913587420669,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2353.2880830025533,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 102.97233299934305,
"inferenceMs": 89.9833330004185,
"imagesPerSecond": 11.113169146505712,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2353.2880830025533,
"totalMs": 2456.2604160018964,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454424064
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.44855859894596506,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 11.08198175114412,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2353.2880830025533,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 99.00679099882836,
"inferenceMs": 90.72304199798964,
"imagesPerSecond": 11.022558084220316,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2353.2880830025533,
"totalMs": 2452.2948740013817,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454424064
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.5477229303578497,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 11.01565926208967,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2353.2880830025533,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 94.52604199759662,
"inferenceMs": 86.455874999956,
"imagesPerSecond": 11.566593941713146,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2353.2880830025533,
"totalMs": 2447.81412500015,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454424064
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0007051299544400536,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 3.0545387309794405,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2353.2880830025533,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 89.57641699817032,
"inferenceMs": 87.24258300208021,
"imagesPerSecond": 11.462292444690181,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
14,
18
]
],
"validPatchCount": 252,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2353.2880830025533,
"totalMs": 2442.8645000007236,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454424064
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.05592406079722423,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.022874748217708,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2353.2880830025533,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 92.17670900034136,
"inferenceMs": 87.78354199966998,
"imagesPerSecond": 11.391656991965071,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
11,
22
]
],
"validPatchCount": 242,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2353.2880830025533,
"totalMs": 2445.4647920028947,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454424064
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.020739834875712404,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.078710606706711,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2353.2880830025533,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 93.57812500093132,
"inferenceMs": 89.96262499931618,
"imagesPerSecond": 11.11572722569624,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
22,
11
]
],
"validPatchCount": 242,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2353.2880830025533,
"totalMs": 2446.8662080034846,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454424064
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07378945255550207,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.4808044991854,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2353.2880830025533,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 102.56800000206567,
"inferenceMs": 91.59804099908797,
"imagesPerSecond": 10.917264049456657,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2455.856083004619,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454424064
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07378945255550207,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.4808044991854,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2353.2880830025533,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 99.25508400192484,
"inferenceMs": 87.98362500238,
"imagesPerSecond": 11.365751297164097,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2452.543167004478,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454424064
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07378945255550207,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.4808044991854,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2353.2880830025533,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 102.82329199981177,
"inferenceMs": 91.71458299897495,
"imagesPerSecond": 10.90339144878603,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2456.111375002365,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454424064
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07378945255550207,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.4808044991854,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2353.2880830025533,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 101.69416600183467,
"inferenceMs": 90.29033399929176,
"imagesPerSecond": 11.075382665079564,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2454.982249004388,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454424064
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07378945255550207,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.4808044991854,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2353.2880830025533,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 97.03379200072959,
"inferenceMs": 85.5674580016057,
"imagesPerSecond": 11.686685842428961,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2450.321875003283,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454424064
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07378945255550207,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.4808044991854,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2384.566750002705,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 175.33750000075088,
"inferenceMs": 156.63908300120966,
"imagesPerSecond": 6.384102746517467,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2384.566750002705,
"totalMs": 2559.9042500034557,
"initializationReused": false,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453146112
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07365939557166712,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.984287222929037,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2384.566750002705,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 99.59008299847483,
"inferenceMs": 88.79391699883854,
"imagesPerSecond": 11.262032736015918,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2384.566750002705,
"totalMs": 2484.1568330011796,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453146112
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.00565307897772982,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.457427054963535,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2384.566750002705,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 101.09637499772361,
"inferenceMs": 88.42741699845647,
"imagesPerSecond": 11.308709831674213,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2384.566750002705,
"totalMs": 2485.6631250004284,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453146112
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.011110889799056167,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.53913587420669,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2384.566750002705,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 98.96541600028286,
"inferenceMs": 86.56229100233759,
"imagesPerSecond": 11.552374462605146,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2384.566750002705,
"totalMs": 2483.5321660029877,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453146112
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.44855859894596506,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 11.08198175114412,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2384.566750002705,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 92.13833299872931,
"inferenceMs": 83.97658400281216,
"imagesPerSecond": 11.90808142382301,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2384.566750002705,
"totalMs": 2476.705083001434,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453146112
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.5477229303578497,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 11.01565926208967,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2384.566750002705,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 93.33658300238312,
"inferenceMs": 85.47225000074832,
"imagesPerSecond": 11.69970370490124,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2384.566750002705,
"totalMs": 2477.903333005088,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453146112
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0007051299544400536,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 3.0545387309794405,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2384.566750002705,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 88.0352079984732,
"inferenceMs": 85.7358329994895,
"imagesPerSecond": 11.663734578819037,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
14,
18
]
],
"validPatchCount": 252,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2384.566750002705,
"totalMs": 2472.601958001178,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453146112
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.05592406079722423,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.022874748217708,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2384.566750002705,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 90.84579200134613,
"inferenceMs": 86.58904199910467,
"imagesPerSecond": 11.548805448273004,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
11,
22
]
],
"validPatchCount": 242,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2384.566750002705,
"totalMs": 2475.412542004051,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453146112
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.020739834875712404,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.078710606706711,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2384.566750002705,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 88.08629200211726,
"inferenceMs": 84.84387499993318,
"imagesPerSecond": 11.786354642580712,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
22,
11
]
],
"validPatchCount": 242,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2384.566750002705,
"totalMs": 2472.653042004822,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453146112
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07378945255550207,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.4808044991854,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2384.566750002705,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 107.63274999771966,
"inferenceMs": 96.9446249982866,
"imagesPerSecond": 10.3151670349715,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2492.1995000004245,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453146112
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07378945255550207,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.4808044991854,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2384.566750002705,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 96.85591599918553,
"inferenceMs": 85.65508399988175,
"imagesPerSecond": 11.674730247201445,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2481.4226660018903,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453146112
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07378945255550207,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.4808044991854,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2384.566750002705,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 102.77133400086313,
"inferenceMs": 91.89854199939873,
"imagesPerSecond": 10.881565455157522,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2487.338084003568,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453146112
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07378945255550207,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.4808044991854,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2384.566750002705,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 98.67283300263807,
"inferenceMs": 87.4935829997412,
"imagesPerSecond": 11.429409628851959,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2483.239583005343,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453146112
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07378945255550207,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.4808044991854,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2384.566750002705,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 96.34533300049952,
"inferenceMs": 85.80237500063959,
"imagesPerSecond": 11.654689045525206,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2480.9120830032043,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1453146112
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07378945255550207,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.4808044991854,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2311.5702499999315,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 165.271166999446,
"inferenceMs": 146.3837080009398,
"imagesPerSecond": 6.831361315110148,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2311.5702499999315,
"totalMs": 2476.8414169993775,
"initializationReused": false,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454456832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07365939557166712,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.984287222929037,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2311.5702499999315,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 96.95329199894331,
"inferenceMs": 85.8154170018679,
"imagesPerSecond": 11.65291779655669,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2311.5702499999315,
"totalMs": 2408.523541998875,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454456832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.00565307897772982,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.457427054963535,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2311.5702499999315,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 105.04945800130372,
"inferenceMs": 91.64154199970653,
"imagesPerSecond": 10.91208177185847,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2311.5702499999315,
"totalMs": 2416.6197080012353,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454456832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.011110889799056167,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.53913587420669,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2311.5702499999315,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 119.88245899920003,
"inferenceMs": 106.90345799957868,
"imagesPerSecond": 9.354234359789757,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2311.5702499999315,
"totalMs": 2431.4527089991316,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454456832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.44855859894596506,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 11.08198175114412,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2311.5702499999315,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 97.67595800076379,
"inferenceMs": 89.65862499826471,
"imagesPerSecond": 11.153416640276989,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2311.5702499999315,
"totalMs": 2409.2462080006953,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454456832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.5477229303578497,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 11.01565926208967,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2311.5702499999315,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 92.97912499823724,
"inferenceMs": 84.88012500311015,
"imagesPerSecond": 11.78132100963987,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2311.5702499999315,
"totalMs": 2404.549374998169,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454456832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0007051299544400536,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 3.0545387309794405,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2311.5702499999315,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 87.44508299787412,
"inferenceMs": 85.05016699928092,
"imagesPerSecond": 11.75776644869439,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
14,
18
]
],
"validPatchCount": 252,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2311.5702499999315,
"totalMs": 2399.0153329978057,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454456832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.05592406079722423,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.022874748217708,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2311.5702499999315,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 89.5009999985632,
"inferenceMs": 85.3512089997821,
"imagesPerSecond": 11.716295664922017,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
11,
22
]
],
"validPatchCount": 242,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2311.5702499999315,
"totalMs": 2401.0712499984947,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454456832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.020739834875712404,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.078710606706711,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2311.5702499999315,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 90.61566700256662,
"inferenceMs": 87.23008400193066,
"imagesPerSecond": 11.463934850479646,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
22,
11
]
],
"validPatchCount": 242,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2311.5702499999315,
"totalMs": 2402.185917002498,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454456832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07378945255550207,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.4808044991854,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2311.5702499999315,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 98.9972499992291,
"inferenceMs": 88.44666699951631,
"imagesPerSecond": 11.306248544170339,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2410.5674999991606,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454456832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07378945255550207,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.4808044991854,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2311.5702499999315,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 100.90029099956155,
"inferenceMs": 90.16787500149803,
"imagesPerSecond": 11.09042438876802,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2412.470540999493,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454456832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07378945255550207,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.4808044991854,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2311.5702499999315,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 98.97825000007288,
"inferenceMs": 88.17133300181013,
"imagesPerSecond": 11.341554742962437,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2410.5485000000044,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454456832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07378945255550207,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.4808044991854,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2311.5702499999315,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 97.60870800164412,
"inferenceMs": 86.73395799996797,
"imagesPerSecond": 11.529509583782275,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2409.1789580015757,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454456832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07378945255550207,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.4808044991854,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2311.5702499999315,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 96.22174999822164,
"inferenceMs": 85.10391600066214,
"imagesPerSecond": 11.750340607031756,
"outputPlacement": "cpu",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2407.791999998153,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454456832
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 554.9197500004084,
"inferenceMs": 526.223958000628,
"imagesPerSecond": 1.9003315694699072,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573224448,
"processInitMs": 3474.3729580004583,
"totalMs": 4029.2927080008667,
"initializationReused": false,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000001952052116394043,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07366134648691514,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.984273881407018,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 58.53841599673615,
"inferenceMs": 45.98233299839194,
"imagesPerSecond": 21.747482887285674,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573224448,
"processInitMs": 3474.3729580004583,
"totalMs": 3532.9113739971945,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 3.166496753692627e-8,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0056531105803969695,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.457413708754423,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 62.43483299840591,
"inferenceMs": 47.874874999251915,
"imagesPerSecond": 20.887782997148836,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573224448,
"processInitMs": 3474.3729580004583,
"totalMs": 3536.807790998864,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 2.300366759300232e-7,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.011111119630641042,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.53912249777645,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.3736250003567,
"inferenceMs": 40.61954200005857,
"imagesPerSecond": 24.618692155577676,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573224448,
"processInitMs": 3474.3729580004583,
"totalMs": 3529.746583000815,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000003248453140258789,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.4485618469689143,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 11.081960514923802,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 63.94616600300651,
"inferenceMs": 54.326624998793704,
"imagesPerSecond": 18.407180641576844,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"processInitMs": 3474.3729580004583,
"totalMs": 3538.319124003465,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 4.172325134277344e-7,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.54772251265058,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 11.01563373641369,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 43.5094160020526,
"inferenceMs": 34.035583998047514,
"imagesPerSecond": 29.381014883051986,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"processInitMs": 3474.3729580004583,
"totalMs": 3517.882374002511,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 2.7939677238464355e-9,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0007051331485854462,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 3.054553971595227,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 33.62145899882307,
"inferenceMs": 29.164249997847946,
"imagesPerSecond": 34.288555339972426,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
14,
18
]
],
"validPatchCount": 252,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"processInitMs": 3474.3729580004583,
"totalMs": 3507.9944169992814,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 7.040798664093018e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.055923356438142946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.022822341595482,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 34.79591599898413,
"inferenceMs": 28.799416999390814,
"imagesPerSecond": 34.72292512105897,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
11,
22
]
],
"validPatchCount": 242,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"processInitMs": 3474.3729580004583,
"totalMs": 3509.1688739994424,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 4.079192876815796e-7,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.020740242880492588,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.078735459751171,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 33.25116700216313,
"inferenceMs": 28.50054200098384,
"imagesPerSecond": 35.087052027483544,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
22,
11
]
],
"validPatchCount": 242,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"processInitMs": 3474.3729580004583,
"totalMs": 3507.6241250026214,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.14699999912409,
"inferenceMs": 42.84274999736226,
"imagesPerSecond": 23.341172078392912,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3529.5199579995824,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.16591600058018,
"inferenceMs": 40.68220800036215,
"imagesPerSecond": 24.580770050413637,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3527.5388740010385,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.730542000877904,
"inferenceMs": 44.49000000022352,
"imagesPerSecond": 22.47696111474435,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3531.103500001336,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.262250000145286,
"inferenceMs": 42.55354099950637,
"imagesPerSecond": 23.49980698460794,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3530.6352080006036,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.67300000216346,
"inferenceMs": 41.03608399964287,
"imagesPerSecond": 24.36879698386188,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3528.045958002622,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.36883400034276,
"inferenceMs": 41.8238329984888,
"imagesPerSecond": 23.90981237984889,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3528.741792000801,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.79037500117556,
"inferenceMs": 42.31266699935077,
"imagesPerSecond": 23.63358471389534,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3529.163333001634,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 47.89600000003702,
"inferenceMs": 32.64037500048289,
"imagesPerSecond": 30.636902915031023,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3522.2689580004953,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 50.85920900091878,
"inferenceMs": 38.222915998630924,
"imagesPerSecond": 26.162315822158053,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3525.232167001377,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.78854200130445,
"inferenceMs": 38.36037500150269,
"imagesPerSecond": 26.068566846930644,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3526.1615000017628,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.52487500204006,
"inferenceMs": 40.5794580001384,
"imagesPerSecond": 24.643010263877585,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3527.8978330024984,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 52.86316699857707,
"inferenceMs": 39.945124997757375,
"imagesPerSecond": 25.034343992067683,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3527.2361249990354,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.408874999557156,
"inferenceMs": 43.0688749984256,
"imagesPerSecond": 23.218623658884876,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3529.7818330000155,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.320541999302804,
"inferenceMs": 38.92133300178102,
"imagesPerSecond": 25.692850754989312,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3525.693499999761,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.46462499862537,
"inferenceMs": 41.71891699661501,
"imagesPerSecond": 23.96994150354234,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3530.8375829990837,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 93.77979200144182,
"inferenceMs": 81.31420900099329,
"imagesPerSecond": 12.29797365412218,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3568.1527500019,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 52.97137500019744,
"inferenceMs": 40.054917000816204,
"imagesPerSecond": 24.965723932959914,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3527.3443330006558,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 58.568874999764375,
"inferenceMs": 45.79133299921523,
"imagesPerSecond": 21.83819370397315,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3532.9418330002227,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.36387500271667,
"inferenceMs": 44.057957999029895,
"imagesPerSecond": 22.697375126237553,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3530.736833003175,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 60.33620800008066,
"inferenceMs": 45.9681249994901,
"imagesPerSecond": 21.75420468011459,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3534.709166000539,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.04637499709497,
"inferenceMs": 43.1563330021163,
"imagesPerSecond": 23.171570206184153,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3530.4193329975533,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.12941699917428,
"inferenceMs": 38.93258400057675,
"imagesPerSecond": 25.68542586295289,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3525.5023749996326,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.471084000193514,
"inferenceMs": 40.588917003333336,
"imagesPerSecond": 24.63726735842387,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3527.844042000652,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.5097920007247,
"inferenceMs": 43.71795800034306,
"imagesPerSecond": 22.873895436565288,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3530.882750001183,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.48670899972785,
"inferenceMs": 42.08483299953514,
"imagesPerSecond": 23.761529480491127,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3530.859667000186,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.84229200010304,
"inferenceMs": 42.3349579978094,
"imagesPerSecond": 23.621140714294427,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3529.2152500005614,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.266916999040404,
"inferenceMs": 41.389917001652066,
"imagesPerSecond": 24.16047367188693,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3528.6398749994987,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 123.1290830000944,
"inferenceMs": 109.49112500020419,
"imagesPerSecond": 9.133160336037601,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3597.5020410005527,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 123.5734580004646,
"inferenceMs": 107.62975000034203,
"imagesPerSecond": 9.291111425947028,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3597.946416000923,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 3474.3729580004583,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 59.72775000191177,
"inferenceMs": 46.57562499778578,
"imagesPerSecond": 21.47045799272775,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 3534.10070800237,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 962052096
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2531.2535829980334,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 190.40850000237697,
"inferenceMs": 162.56629199779127,
"imagesPerSecond": 6.1513367113865565,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573240832,
"processInitMs": 2531.2535829980334,
"totalMs": 2721.6620830004103,
"initializationReused": false,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 813858816
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000001952052116394043,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07366134648691514,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.984273881407018,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2531.2535829980334,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.32170799758751,
"inferenceMs": 39.2258749998291,
"imagesPerSecond": 25.493376502228614,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573240832,
"processInitMs": 2531.2535829980334,
"totalMs": 2585.575290995621,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813858816
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 3.166496753692627e-8,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0056531105803969695,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.457413708754423,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2531.2535829980334,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 61.59391699839034,
"inferenceMs": 46.781708999333205,
"imagesPerSecond": 21.375875772607053,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573240832,
"processInitMs": 2531.2535829980334,
"totalMs": 2592.8474999964237,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813858816
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 2.300366759300232e-7,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.011111119630641042,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.53912249777645,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2531.2535829980334,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.72333300005994,
"inferenceMs": 40.60104099698947,
"imagesPerSecond": 24.629910353139692,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573240832,
"processInitMs": 2531.2535829980334,
"totalMs": 2586.9769159980933,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813858816
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000003248453140258789,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.4485618469689143,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 11.081960514923802,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2531.2535829980334,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 50.332875001913635,
"inferenceMs": 40.677666998817585,
"imagesPerSecond": 24.58351409457843,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579565056,
"processInitMs": 2531.2535829980334,
"totalMs": 2581.586457999947,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813858816
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 4.172325134277344e-7,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.54772251265058,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 11.01563373641369,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2531.2535829980334,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.46041700209025,
"inferenceMs": 42.38670899940189,
"imagesPerSecond": 23.59230106810394,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579565056,
"processInitMs": 2531.2535829980334,
"totalMs": 2582.7140000001236,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813858816
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 2.7939677238464355e-9,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0007051331485854462,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 3.054553971595227,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2531.2535829980334,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 32.81283300020732,
"inferenceMs": 28.74379100103397,
"imagesPerSecond": 34.79012215069432,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
14,
18
]
],
"validPatchCount": 252,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"processInitMs": 2531.2535829980334,
"totalMs": 2564.0664159982407,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813858816
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 7.040798664093018e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.055923356438142946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.022822341595482,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2531.2535829980334,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 36.36291599832475,
"inferenceMs": 30.33487500215415,
"imagesPerSecond": 32.96535752756481,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
11,
22
]
],
"validPatchCount": 242,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"processInitMs": 2531.2535829980334,
"totalMs": 2567.616498996358,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813858816
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 4.079192876815796e-7,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.020740242880492588,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.078735459751171,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2531.2535829980334,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 33.74333300234866,
"inferenceMs": 28.480832999775885,
"imagesPerSecond": 35.1113325936734,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
22,
11
]
],
"validPatchCount": 242,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"processInitMs": 2531.2535829980334,
"totalMs": 2564.996916000382,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813858816
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2531.2535829980334,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.17541700100992,
"inferenceMs": 42.847333003010135,
"imagesPerSecond": 23.338675476715142,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 2586.4289999990433,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813858816
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2531.2535829980334,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.95108299914864,
"inferenceMs": 44.627250001212815,
"imagesPerSecond": 22.40783377808006,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 2588.204665997182,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813858816
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2531.2535829980334,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.27458400299656,
"inferenceMs": 40.49900000245543,
"imagesPerSecond": 24.691967701409183,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 2585.52816700103,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813858816
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2531.2535829980334,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.9501249982568,
"inferenceMs": 37.5567079972825,
"imagesPerSecond": 26.626401868671696,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 2583.20370799629,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813858816
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2531.2535829980334,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.06820799928391,
"inferenceMs": 40.34470899932785,
"imagesPerSecond": 24.786397641798835,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 2586.3217909973173,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813858816
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2559.1396250019898,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 181.14991699985694,
"inferenceMs": 153.87004200238152,
"imagesPerSecond": 6.498990882088162,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573240832,
"processInitMs": 2559.1396250019898,
"totalMs": 2740.2895420018467,
"initializationReused": false,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 813449216
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000001952052116394043,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07366134648691514,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.984273881407018,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2559.1396250019898,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 49.8263749977923,
"inferenceMs": 37.201916999038076,
"imagesPerSecond": 26.880335226430855,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573240832,
"processInitMs": 2559.1396250019898,
"totalMs": 2608.965999999782,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813449216
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 3.166496753692627e-8,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0056531105803969695,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.457413708754423,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2559.1396250019898,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 58.34908399992855,
"inferenceMs": 43.61150000113412,
"imagesPerSecond": 22.92973183618988,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573240832,
"processInitMs": 2559.1396250019898,
"totalMs": 2617.4887090019183,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813449216
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 2.300366759300232e-7,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.011111119630641042,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.53912249777645,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2559.1396250019898,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 52.31691599692567,
"inferenceMs": 37.384374998509884,
"imagesPerSecond": 26.74914319257335,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573240832,
"processInitMs": 2559.1396250019898,
"totalMs": 2611.4565409989154,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813449216
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000003248453140258789,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.4485618469689143,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 11.081960514923802,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2559.1396250019898,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 61.29437499839696,
"inferenceMs": 51.55137499968987,
"imagesPerSecond": 19.39812468641265,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579565056,
"processInitMs": 2559.1396250019898,
"totalMs": 2620.4340000003867,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 813449216
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 4.172325134277344e-7,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.54772251265058,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 11.01563373641369,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2559.1396250019898,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 44.566415999724995,
"inferenceMs": 34.73708299861755,
"imagesPerSecond": 28.78767915083133,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579565056,
"processInitMs": 2559.1396250019898,
"totalMs": 2603.7060410017148,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813449216
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 2.7939677238464355e-9,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0007051331485854462,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 3.054553971595227,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2559.1396250019898,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 34.02404100052081,
"inferenceMs": 29.973874999996042,
"imagesPerSecond": 33.36238641150442,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
14,
18
]
],
"validPatchCount": 252,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"processInitMs": 2559.1396250019898,
"totalMs": 2593.1636660025106,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813449216
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 7.040798664093018e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.055923356438142946,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.022822341595482,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2559.1396250019898,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 36.07304099932662,
"inferenceMs": 30.346374998771353,
"imagesPerSecond": 32.95286504699449,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
11,
22
]
],
"validPatchCount": 242,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"processInitMs": 2559.1396250019898,
"totalMs": 2595.2126660013164,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813449216
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 4.079192876815796e-7,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.020740242880492588,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.078735459751171,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2559.1396250019898,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 33.39445899837301,
"inferenceMs": 28.627833002246916,
"imagesPerSecond": 34.9310407085829,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
22,
11
]
],
"validPatchCount": 242,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"processInitMs": 2559.1396250019898,
"totalMs": 2592.534084000363,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813449216
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2559.1396250019898,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 57.10583300242433,
"inferenceMs": 44.210000000020955,
"imagesPerSecond": 22.619316896619,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 2616.245458004414,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813449216
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2559.1396250019898,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.168250000453554,
"inferenceMs": 40.317625000170665,
"imagesPerSecond": 24.80304829453042,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 2612.3078750024433,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813449216
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2559.1396250019898,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 58.36987500151736,
"inferenceMs": 45.628583000507206,
"imagesPerSecond": 21.916087115589015,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 2617.509500003507,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813449216
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2559.1396250019898,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 58.5264589972212,
"inferenceMs": 43.82712499864283,
"imagesPerSecond": 22.81691988764872,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 2617.666083999211,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813449216
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.419156074523926e-7,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.0737902942346409,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.48080250783746,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2559.1396250019898,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.824457998009166,
"inferenceMs": 38.641250001091976,
"imagesPerSecond": 25.87907999797472,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1579548672,
"totalMs": 2610.964082999999,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 813449216
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002214781939983368,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07157475594976859,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.453124982639945,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2447.737375001452,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 489.57099999825004,
"inferenceMs": 461.34504200017545,
"imagesPerSecond": 2.1675750446227178,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2447.737375001452,
"totalMs": 2937.308374999702,
"initializationReused": false,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1413644288
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.001574300229549408,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07208509384508943,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.960937536344309,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2447.737375001452,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.14658299903385,
"inferenceMs": 40.448624997225124,
"imagesPerSecond": 24.722719253586554,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2447.737375001452,
"totalMs": 2500.883958000486,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1413644288
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00010143732652068138,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.00555168695723296,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.453125033530517,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2447.737375001452,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.53220799952396,
"inferenceMs": 39.84333300104481,
"imagesPerSecond": 25.098301890902977,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2447.737375001452,
"totalMs": 2501.269583000976,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1413644288
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00004116259515285492,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.011152029565892008,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.5390625528161666,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2447.737375001452,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 49.703417000273475,
"inferenceMs": 34.30599999774131,
"imagesPerSecond": 29.149419928462645,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2447.737375001452,
"totalMs": 2497.4407920017256,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1413644288
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.003041595220565796,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.4455169135653705,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 11.062499939860901,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2447.737375001452,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.266332997736754,
"inferenceMs": 43.51079200205277,
"imagesPerSecond": 22.982803897314064,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"processInitMs": 2447.737375001452,
"totalMs": 2501.003707999189,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1413644288
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.008759617805480957,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.5389633924569353,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 10.984375098656317,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2447.737375001452,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 43.683250001777196,
"inferenceMs": 31.97737500158837,
"imagesPerSecond": 31.27211035772412,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"processInitMs": 2447.737375001452,
"totalMs": 2491.4206250032294,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1413644288
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000013556040357798338,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0006916793645359576,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 3.0390624231121164,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2447.737375001452,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 27.358457999071106,
"inferenceMs": 23.563290997117292,
"imagesPerSecond": 42.438893621537794,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
14,
18
]
],
"validPatchCount": 252,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"processInitMs": 2447.737375001452,
"totalMs": 2475.0958330005233,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1413644288
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0009259209036827087,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.05499810262926985,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.000000059461714,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2447.737375001452,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 30.19362499981071,
"inferenceMs": 24.59883299889043,
"imagesPerSecond": 40.65233501301085,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
11,
22
]
],
"validPatchCount": 242,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"processInitMs": 2447.737375001452,
"totalMs": 2477.931000001263,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1413644288
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002691466361284256,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.020470636623940663,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.062500019456872,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2447.737375001452,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 27.519041002960876,
"inferenceMs": 22.925375000340864,
"imagesPerSecond": 43.619788116230666,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
22,
11
]
],
"validPatchCount": 242,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"processInitMs": 2447.737375001452,
"totalMs": 2475.256416004413,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1413644288
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002214781939983368,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07157475594976859,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.453124982639945,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2447.737375001452,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 46.830000002955785,
"inferenceMs": 31.428333000803832,
"imagesPerSecond": 31.818423203496774,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2494.567375004408,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1413644288
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002214781939983368,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07157475594976859,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.453124982639945,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2447.737375001452,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 42.60466699997778,
"inferenceMs": 29.709833001106745,
"imagesPerSecond": 33.658889969618755,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2490.34204200143,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1413644288
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002214781939983368,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07157475594976859,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.453124982639945,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2447.737375001452,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 57.19104200034053,
"inferenceMs": 43.36591699757264,
"imagesPerSecond": 23.059583867579093,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2504.9284170017927,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1413644288
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002214781939983368,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07157475594976859,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.453124982639945,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2447.737375001452,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.906916003033984,
"inferenceMs": 36.06150000268826,
"imagesPerSecond": 27.730405000497854,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2496.644291004486,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1413644288
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002214781939983368,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07157475594976859,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.453124982639945,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2447.737375001452,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 50.57954199946835,
"inferenceMs": 38.50879200035706,
"imagesPerSecond": 25.96809580499767,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2498.3169170009205,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1413644288
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002214781939983368,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07157475594976859,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.453124982639945,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2666.498082999169,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 227.94529200109537,
"inferenceMs": 199.17091699971934,
"imagesPerSecond": 5.020813355001067,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2666.498082999169,
"totalMs": 2894.4433750002645,
"initializationReused": false,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1436827648
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.001574300229549408,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07208509384508943,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.960937536344309,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2666.498082999169,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 120.54779100071755,
"inferenceMs": 107.03783399731037,
"imagesPerSecond": 9.342490992718778,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2666.498082999169,
"totalMs": 2787.0458739998867,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1436827648
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00010143732652068138,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.00555168695723296,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.453125033530517,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2666.498082999169,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 118.53195799994865,
"inferenceMs": 101.31845800060546,
"imagesPerSecond": 9.869869910515458,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2666.498082999169,
"totalMs": 2785.030040999118,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1436827648
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00004116259515285492,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.011152029565892008,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.5390625528161666,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2666.498082999169,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 113.55895800079452,
"inferenceMs": 98.19129099923884,
"imagesPerSecond": 10.184202588880838,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2666.498082999169,
"totalMs": 2780.0570409999636,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1436827648
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.003041595220565796,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.4455169135653705,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 11.062499939860901,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2666.498082999169,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 116.88900000081048,
"inferenceMs": 107.06637499970384,
"imagesPerSecond": 9.340000537075866,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"processInitMs": 2666.498082999169,
"totalMs": 2783.3870829999796,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1436827648
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.008759617805480957,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.5389633924569353,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 10.984375098656317,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2666.498082999169,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 101.78833300233237,
"inferenceMs": 91.90770799978054,
"imagesPerSecond": 10.88048023134673,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"processInitMs": 2666.498082999169,
"totalMs": 2768.2864160015015,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1436827648
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000013556040357798338,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0006916793645359576,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 3.0390624231121164,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2666.498082999169,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 88.14979099770426,
"inferenceMs": 83.84975000080885,
"imagesPerSecond": 11.926093995394782,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
14,
18
]
],
"validPatchCount": 252,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"processInitMs": 2666.498082999169,
"totalMs": 2754.6478739968734,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1436827648
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0009259209036827087,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.05499810262926985,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.000000059461714,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2666.498082999169,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 86.60800000143354,
"inferenceMs": 80.30312499977299,
"imagesPerSecond": 12.452815503790506,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
11,
22
]
],
"validPatchCount": 242,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"processInitMs": 2666.498082999169,
"totalMs": 2753.1060830006027,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1436827648
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002691466361284256,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.020470636623940663,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.062500019456872,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2666.498082999169,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 85.92387499811593,
"inferenceMs": 78.42424999762443,
"imagesPerSecond": 12.751157964918903,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
22,
11
]
],
"validPatchCount": 242,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"processInitMs": 2666.498082999169,
"totalMs": 2752.421957997285,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1436827648
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002214781939983368,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07157475594976859,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.453124982639945,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2666.498082999169,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 100.08683300111443,
"inferenceMs": 84.94833299846505,
"imagesPerSecond": 11.771861373878512,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2766.5849160002836,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1436827648
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002214781939983368,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07157475594976859,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.453124982639945,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2666.498082999169,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 100.69899999871268,
"inferenceMs": 88.2636669994099,
"imagesPerSecond": 11.329690165792517,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2767.197082997882,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1436827648
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002214781939983368,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07157475594976859,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.453124982639945,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2666.498082999169,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 107.69983400314231,
"inferenceMs": 94.40875000291271,
"imagesPerSecond": 10.592238536885064,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2774.1979170023114,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1436827648
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002214781939983368,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07157475594976859,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.453124982639945,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2666.498082999169,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 117.73133300084737,
"inferenceMs": 104.9622080026893,
"imagesPerSecond": 9.52723860357795,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2784.2294160000165,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1436827648
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002214781939983368,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07157475594976859,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.453124982639945,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2666.498082999169,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 113.26966699925833,
"inferenceMs": 99.82904199932818,
"imagesPerSecond": 10.017125076756017,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2779.7677499984275,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1436827648
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002214781939983368,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07157475594976859,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.453124982639945,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2503.4398329989926,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 199.53841700043995,
"inferenceMs": 170.96775000027264,
"imagesPerSecond": 5.849056327865374,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2503.4398329989926,
"totalMs": 2702.9782499994326,
"initializationReused": false,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1527234560
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.001574300229549408,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07208509384508943,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.960937536344309,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2503.4398329989926,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 50.51983399971505,
"inferenceMs": 37.564124999335036,
"imagesPerSecond": 26.62114451002658,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2503.4398329989926,
"totalMs": 2553.9596669987077,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1527234560
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00010143732652068138,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.00555168695723296,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.453125033530517,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2503.4398329989926,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 47.86399999647983,
"inferenceMs": 32.78683299868135,
"imagesPerSecond": 30.500048603054125,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2503.4398329989926,
"totalMs": 2551.3038329954725,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1527234560
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00004116259515285492,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.011152029565892008,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.5390625528161666,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2503.4398329989926,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 50.92791600327473,
"inferenceMs": 37.01641700172331,
"imagesPerSecond": 27.01504037933884,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2503.4398329989926,
"totalMs": 2554.3677490022674,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1527234560
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.003041595220565796,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.4455169135653705,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 11.062499939860901,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2503.4398329989926,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.29879199739662,
"inferenceMs": 41.407832999539096,
"imagesPerSecond": 24.15002011844307,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120288768,
"processInitMs": 2503.4398329989926,
"totalMs": 2554.7386249963893,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1527234560
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.008759617805480957,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.5389633924569353,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 10.984375098656317,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2503.4398329989926,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 46.61183300049743,
"inferenceMs": 35.21620800165692,
"imagesPerSecond": 28.39601583319107,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120288768,
"processInitMs": 2503.4398329989926,
"totalMs": 2550.05166599949,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1527234560
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000013556040357798338,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0006916793645359576,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 3.0390624231121164,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2503.4398329989926,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 27.86062499944819,
"inferenceMs": 23.690624999289867,
"imagesPerSecond": 42.21079013449309,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
14,
18
]
],
"validPatchCount": 252,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120288768,
"processInitMs": 2503.4398329989926,
"totalMs": 2531.300457998441,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1527234560
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0009259209036827087,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.05499810262926985,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.000000059461714,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2503.4398329989926,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 28.998041001614183,
"inferenceMs": 22.745416001271224,
"imagesPerSecond": 43.964902639903826,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
11,
22
]
],
"validPatchCount": 242,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120288768,
"processInitMs": 2503.4398329989926,
"totalMs": 2532.437874000607,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1527234560
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0002691466361284256,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.020470636623940663,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.062500019456872,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2503.4398329989926,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 27.235790999839082,
"inferenceMs": 22.23187499839696,
"imagesPerSecond": 44.98046161523063,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
22,
11
]
],
"validPatchCount": 242,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120288768,
"processInitMs": 2503.4398329989926,
"totalMs": 2530.6756239988317,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1527234560
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002214781939983368,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07157475594976859,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.453124982639945,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2503.4398329989926,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.189749996818136,
"inferenceMs": 40.16541699820664,
"imagesPerSecond": 24.897040158817457,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2556.629582995811,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1527234560
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002214781939983368,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07157475594976859,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.453124982639945,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2503.4398329989926,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 50.82241700074519,
"inferenceMs": 36.98274999987916,
"imagesPerSecond": 27.039633342660224,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2554.262249999738,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1527234560
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002214781939983368,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07157475594976859,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.453124982639945,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2503.4398329989926,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 50.199082998005906,
"inferenceMs": 37.498792000405956,
"imagesPerSecond": 26.66752571627305,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2553.6389159969985,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1527234560
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002214781939983368,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07157475594976859,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.453124982639945,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2503.4398329989926,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 49.324791001708945,
"inferenceMs": 36.46045899949968,
"imagesPerSecond": 27.426972326753273,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2552.7646240007016,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1527234560
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002214781939983368,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.07157475594976859,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.453124982639945,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2503.4398329989926,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 47.67579200051841,
"inferenceMs": 35.373374998016516,
"imagesPerSecond": 28.269849853345143,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2551.115624999511,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1527234560
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006551489233970642,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.08033969106691075,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.500000036280149,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2464.4118750002235,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 493.2937910016335,
"inferenceMs": 463.01175000189687,
"imagesPerSecond": 2.159772403175304,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2464.4118750002235,
"totalMs": 2957.705666001857,
"initializationReused": false,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1523073024
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006670966744422913,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.08032815807928273,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.999999994460206,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2464.4118750002235,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.06029200073681,
"inferenceMs": 40.28937499970198,
"imagesPerSecond": 24.82043963222058,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2464.4118750002235,
"totalMs": 2517.4721670009603,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1523073024
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.001465066336095333,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.007117205548183847,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.43749994303014,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2464.4118750002235,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.03637499996694,
"inferenceMs": 36.39516600014758,
"imagesPerSecond": 27.47617636902508,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2464.4118750002235,
"totalMs": 2515.4482500001905,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1523073024
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002105953171849251,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.013215859732554236,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.562500109857006,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2464.4118750002235,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 47.11029100144515,
"inferenceMs": 32.468290999531746,
"imagesPerSecond": 30.79928044301506,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2464.4118750002235,
"totalMs": 2511.5221660016687,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1523073024
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.01074761152267456,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.4378096582049693,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 10.93750002175806,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2464.4118750002235,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.801708999846596,
"inferenceMs": 42.451041997992434,
"imagesPerSecond": 23.556547800341185,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"processInitMs": 2464.4118750002235,
"totalMs": 2516.21358400007,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1523073024
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.01653343439102173,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.5311879642904387,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 10.87500023502266,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2464.4118750002235,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 42.03879200213123,
"inferenceMs": 29.93162500206381,
"imagesPerSecond": 33.4094791021553,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"processInitMs": 2464.4118750002235,
"totalMs": 2506.4506670023547,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1523073024
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000303732231259346,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0006696066120639443,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 2.874999854694604,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2464.4118750002235,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 26.747250001790235,
"inferenceMs": 23.174417001428083,
"imagesPerSecond": 43.15103158532,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
14,
18
]
],
"validPatchCount": 252,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"processInitMs": 2464.4118750002235,
"totalMs": 2491.1591250020138,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1523073024
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.007783789187669754,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.06370659790809441,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.000000013087108,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2464.4118750002235,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 29.87441700315685,
"inferenceMs": 23.429415999999037,
"imagesPerSecond": 42.68138821727529,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
11,
22
]
],
"validPatchCount": 242,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"processInitMs": 2464.4118750002235,
"totalMs": 2494.2862920033804,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1523073024
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0022196676582098007,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.02295844375839806,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.124999981901992,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2464.4118750002235,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 28.386958998453338,
"inferenceMs": 23.220499999297317,
"imagesPerSecond": 43.06539480331006,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
22,
11
]
],
"validPatchCount": 242,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"processInitMs": 2464.4118750002235,
"totalMs": 2492.798833998677,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1523073024
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006551489233970642,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.08033969106691075,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.500000036280149,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2464.4118750002235,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 49.32708300111699,
"inferenceMs": 36.39245800150093,
"imagesPerSecond": 27.47822089837288,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2513.7389580013405,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1523073024
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006551489233970642,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.08033969106691075,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.500000036280149,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2464.4118750002235,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 44.194416997925146,
"inferenceMs": 31.48966599837877,
"imagesPerSecond": 31.756449879509187,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2508.6062919981487,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1523073024
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006551489233970642,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.08033969106691075,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.500000036280149,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2464.4118750002235,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 52.37574999773642,
"inferenceMs": 39.66379199846415,
"imagesPerSecond": 25.21191115662168,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2516.78762499796,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1523073024
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006551489233970642,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.08033969106691075,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.500000036280149,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2464.4118750002235,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.70474999977159,
"inferenceMs": 39.55504199984716,
"imagesPerSecond": 25.28122710636646,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2516.116624999995,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1523073024
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006551489233970642,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.08033969106691075,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.500000036280149,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2464.4118750002235,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.16620899873669,
"inferenceMs": 35.95816600136459,
"imagesPerSecond": 27.81009465171418,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2512.57808399896,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1523073024
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006551489233970642,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.08033969106691075,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.500000036280149,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2442.2428329999093,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 193.33916599862278,
"inferenceMs": 164.721749999444,
"imagesPerSecond": 6.070843710702293,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2442.2428329999093,
"totalMs": 2635.581998998532,
"initializationReused": false,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1305329664
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006670966744422913,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.08032815807928273,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.999999994460206,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2442.2428329999093,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 57.342834003065946,
"inferenceMs": 42.83679100262816,
"imagesPerSecond": 23.34441905180636,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2442.2428329999093,
"totalMs": 2499.5856670029752,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1305329664
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.001465066336095333,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.007117205548183847,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.43749994303014,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2442.2428329999093,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 55.36420900170924,
"inferenceMs": 40.548375000071246,
"imagesPerSecond": 24.661900754302557,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2442.2428329999093,
"totalMs": 2497.6070420016185,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1305329664
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002105953171849251,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.013215859732554236,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.562500109857006,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2442.2428329999093,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 49.345291998179164,
"inferenceMs": 34.54220799903851,
"imagesPerSecond": 28.95008912076018,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2442.2428329999093,
"totalMs": 2491.5881249980885,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1305329664
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.01074761152267456,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.4378096582049693,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 10.93750002175806,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2442.2428329999093,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 60.046124999644235,
"inferenceMs": 50.531207998574246,
"imagesPerSecond": 19.78975052463055,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120288768,
"processInitMs": 2442.2428329999093,
"totalMs": 2502.2889579995535,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1305329664
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.01653343439102173,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.5311879642904387,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 10.87500023502266,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2442.2428329999093,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 40.49999999915599,
"inferenceMs": 30.710166996868793,
"imagesPerSecond": 32.56250609454387,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120288768,
"processInitMs": 2442.2428329999093,
"totalMs": 2482.7428329990653,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1305329664
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000303732231259346,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0006696066120639443,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 2.874999854694604,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2442.2428329999093,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 27.16524999777903,
"inferenceMs": 23.018125000817236,
"imagesPerSecond": 43.44402508738205,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
14,
18
]
],
"validPatchCount": 252,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120288768,
"processInitMs": 2442.2428329999093,
"totalMs": 2469.4080829976883,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1305329664
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.007783789187669754,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.06370659790809441,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.000000013087108,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2442.2428329999093,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 29.107541999110254,
"inferenceMs": 23.17429200047627,
"imagesPerSecond": 43.151264339788604,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
11,
22
]
],
"validPatchCount": 242,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120288768,
"processInitMs": 2442.2428329999093,
"totalMs": 2471.3503749990196,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1305329664
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0022196676582098007,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.02295844375839806,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.124999981901992,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2442.2428329999093,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 28.77479100061464,
"inferenceMs": 23.599708001711406,
"imagesPerSecond": 42.37340563397996,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
22,
11
]
],
"validPatchCount": 242,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120288768,
"processInitMs": 2442.2428329999093,
"totalMs": 2471.017624000524,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1305329664
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006551489233970642,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.08033969106691075,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.500000036280149,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2442.2428329999093,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.475957999471575,
"inferenceMs": 36.26645799886319,
"imagesPerSecond": 27.57368806270924,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2490.718790999381,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1305329664
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006551489233970642,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.08033969106691075,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.500000036280149,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2442.2428329999093,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.165082998224534,
"inferenceMs": 39.757541999279056,
"imagesPerSecond": 25.15246038143237,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2496.407915998134,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1305329664
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006551489233970642,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.08033969106691075,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.500000036280149,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2442.2428329999093,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 82.57533299911302,
"inferenceMs": 69.96283300031791,
"imagesPerSecond": 14.293303417193755,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2524.8181659990223,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1305329664
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006551489233970642,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.08033969106691075,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.500000036280149,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2442.2428329999093,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 50.04954199830536,
"inferenceMs": 35.70229099932476,
"imagesPerSecond": 28.00940701589467,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2492.2923749982147,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1305329664
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006551489233970642,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.08033969106691075,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.500000036280149,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2442.2428329999093,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 42.36795900214929,
"inferenceMs": 28.787708000891143,
"imagesPerSecond": 34.73704818629688,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2484.6107920020586,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1305329664
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006551489233970642,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.08033969106691075,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.500000036280149,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2584.5574580016546,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 213.0518750018382,
"inferenceMs": 182.9645839970908,
"imagesPerSecond": 5.465538620391694,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2584.5574580016546,
"totalMs": 2797.609333003493,
"initializationReused": false,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1528954880
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006670966744422913,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.08032815807928273,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.999999994460206,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2584.5574580016546,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.008458999102004,
"inferenceMs": 39.80841600059648,
"imagesPerSecond": 25.120316266415028,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2584.5574580016546,
"totalMs": 2637.5659170007566,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1528954880
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.001465066336095333,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.007117205548183847,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.43749994303014,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2584.5574580016546,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 58.42416700033937,
"inferenceMs": 43.364291999751,
"imagesPerSecond": 23.060447983463952,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2584.5574580016546,
"totalMs": 2642.981625001994,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1528954880
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002105953171849251,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.013215859732554236,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.562500109857006,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2584.5574580016546,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.35579100230825,
"inferenceMs": 38.701457997376565,
"imagesPerSecond": 25.838819820891153,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
13,
19
]
],
"validPatchCount": 247,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1113964544,
"processInitMs": 2584.5574580016546,
"totalMs": 2638.913249003963,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1528954880
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.01074761152267456,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.4378096582049693,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 10.93750002175806,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2584.5574580016546,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 62.13512499743956,
"inferenceMs": 52.38479200124857,
"imagesPerSecond": 19.089509794677916,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120288768,
"processInitMs": 2584.5574580016546,
"totalMs": 2646.692582999094,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1528954880
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.01653343439102173,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.5311879642904387,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 10.87500023502266,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2584.5574580016546,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 76.23316600074759,
"inferenceMs": 59.263541999825975,
"imagesPerSecond": 16.873780510839808,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
16,
16
]
],
"validPatchCount": 256,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120288768,
"processInitMs": 2584.5574580016546,
"totalMs": 2660.790624002402,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1528954880
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000303732231259346,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0006696066120639443,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 2.874999854694604,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2584.5574580016546,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 46.712750001461245,
"inferenceMs": 42.41987499699462,
"imagesPerSecond": 23.57385541732144,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
14,
18
]
],
"validPatchCount": 252,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120288768,
"processInitMs": 2584.5574580016546,
"totalMs": 2631.270208003116,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1528954880
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.007783789187669754,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.06370659790809441,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.000000013087108,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2584.5574580016546,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 56.65475000205333,
"inferenceMs": 49.177625001902925,
"imagesPerSecond": 20.334450880076155,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
11,
22
]
],
"validPatchCount": 242,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"processInitMs": 2584.5574580016546,
"totalMs": 2641.212208003708,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1528954880
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0022196676582098007,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.02295844375839806,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 7.124999981901992,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2584.5574580016546,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 64.22429200029,
"inferenceMs": 56.30083299911348,
"imagesPerSecond": 17.76172654524927,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
22,
11
]
],
"validPatchCount": 242,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"processInitMs": 2584.5574580016546,
"totalMs": 2648.7817500019446,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1528954880
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006551489233970642,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.08033969106691075,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.500000036280149,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2584.5574580016546,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.25191600018297,
"inferenceMs": 36.495749998721294,
"imagesPerSecond": 27.400450738374662,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2635.8093740018376,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1528954880
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006551489233970642,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.08033969106691075,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.500000036280149,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2584.5574580016546,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.42537500069011,
"inferenceMs": 36.38020799917285,
"imagesPerSecond": 27.487473409243183,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2635.9828330023447,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1528954880
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006551489233970642,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.08033969106691075,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.500000036280149,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2584.5574580016546,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 50.660500000958564,
"inferenceMs": 36.26087500015274,
"imagesPerSecond": 27.57793351638061,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2635.217958002613,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1528954880
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006551489233970642,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.08033969106691075,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.500000036280149,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2584.5574580016546,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 52.084708000620594,
"inferenceMs": 39.56349999862141,
"imagesPerSecond": 25.27582241295247,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2636.642166002275,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1528954880
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.006551489233970642,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.08033969106691075,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.500000036280149,
"nativeAspectRatioGridAndMask": true,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip2",
"patchSize": 16,
"processorType": "Siglip2ImageProcessorPil",
"loadMs": 2584.5574580016546,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 114.35645899837255,
"inferenceMs": 100.28012499969918,
"imagesPerSecond": 9.972065750845442,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
256,
768
],
"spatialShapes": [
[
15,
17
]
],
"validPatchCount": 255,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1120272384,
"totalMs": 2698.913917000027,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "b53b807d3a2d5e2b3911292f2d69e5341cdc064c",
"checkpoint": "siglip2-base-patch16-naflex",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1528954880
}
}
]SigLIP 2 Base · Patch32 · 256 — Image & Text Matching Desktop
235 configurations testedQuality criteria
{ "minimumTop1Accuracy": 1, "maximumCpuScoreDifference": 0.03, "finiteScoresInUnitInterval": true, "scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification", "performanceTargets": { "goodInferenceMs": 250, "acceptableInferenceMs": 1000, "description": "One image at pinned256px FixRes or256patch NaFlex default and four labels; good<=250ms,acceptable<=1000ms" }, "version": 2, "minimumLogOddsMargin": 1, "gateRevisionReason": "V1 required0.05absolute sigmoid margin, which conflated calibrated confidence with ranking despite the native independent sigmoid contract. All six semantic fixtures ranked correctly but dog/portrait failed absolute margin. Preserve fullv1failures;v2requires expected top1plus ≥1natural-log-odds separation (odds ratio≥e), finite scores and unchanged0.03CPU agreement. This measures ranking separation without treating sigmoid scores as calibrated probabilities." }
| Precision | Device | Result | Platform | Inference / load | |
|---|---|---|---|---|---|
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 17 ms / 2612 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 17 ms / 2612 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 17 ms / 2488 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 17 ms / 2488 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 2488 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 2468 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 18 ms / 2468 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 19 ms / 2671 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 19 ms / 2541 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 19 ms / 2468 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 19 ms / 2568 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 19 ms / 2541 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 19 ms / 2568 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 19 ms / 2541 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 20 ms / 2683 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 20 ms / 2683 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 22 ms / 2671 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 22 ms / 2568 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 2612 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 23 ms / 2671 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 27 ms / 2612 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 28 ms / 2683 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 28 ms / 2488 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 30 ms / 2612 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 31 ms / 2468 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 32 ms / 2568 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 32 ms / 2612 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 32 ms / 2488 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 32 ms / 2568 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 32 ms / 2683 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 32 ms / 2671 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 32 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 32 ms / 2541 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 32 ms / 2612 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 33 ms / 2541 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 33 ms / 2683 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 33 ms / 2568 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 33 ms / 2683 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 33 ms / 2468 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 33 ms / 2683 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 2568 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 2468 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 2671 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 2488 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 2468 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 2671 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 2671 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 2488 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 34 ms / 2568 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 2683 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 2568 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 2488 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 2671 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 2922 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 2488 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 2541 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 2488 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 35 ms / 2468 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 2671 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 2671 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 2683 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 2541 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 36 ms / 2468 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 37 ms / 2541 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 37 ms / 2541 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 37 ms / 2541 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 37 ms / 2683 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 38 ms / 2468 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 38 ms / 2488 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 38 ms / 2568 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 38 ms / 2468 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 38 ms / 2612 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 38 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 39 ms / 2683 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 39 ms / 2488 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 39 ms / 2683 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 39 ms / 2468 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 39 ms / 2541 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 2671 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 2568 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 40 ms / 2568 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 2488 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 41 ms / 2671 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 42 ms / 2683 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 2568 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 43 ms / 2671 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 47 ms / 2922 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 48 ms / 2560 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 48 ms / 2560 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 51 ms / 2560 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 56 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 57 ms / 2541 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 59 ms / 2560 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 59 ms / 2560 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 61 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 61 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 61 ms / 2541 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 62 ms / 2560 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 62 ms / 2560 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 62 ms / 2560 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 63 ms / 2612 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 63 ms / 2560 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 64 ms / 2612 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 64 ms / 2612 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 65 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 65 ms / 2922 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 65 ms / 2560 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 65 ms / 2922 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 65 ms / 2560 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 66 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 66 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 66 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 66 ms / 2922 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 66 ms / 2612 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 66 ms / 2560 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 66 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 67 ms / 2922 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 67 ms / 2612 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 68 ms / 2468 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 68 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 68 ms / 2922 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 68 ms / 2560 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 68 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 69 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 69 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 69 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 69 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 69 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 69 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 70 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 70 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 70 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 70 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 70 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 70 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 70 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 71 ms / 2922 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 73 ms / 2922 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 117 ms / 2468 ms |
BF16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 118 ms / 2612 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 126 ms / 2683 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 128 ms / 2568 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 138 ms / 2541 ms |
BF16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 138 ms / 2560 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 226 ms / 2922 ms |
FP16 | mpsAccelerated | Passed | Moderate | macos · aarch64 | 390 ms / 2488 ms |
BF16 | mpsAccelerated | Passed | Moderate | macos · aarch64 | 437 ms / 2671 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 61 ms / 2402 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 61 ms / 2402 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 61 ms / 2320 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 61 ms / 2364 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 61 ms / 2320 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 61 ms / 2402 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 62 ms / 2364 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 62 ms / 2320 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 62 ms / 2364 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 62 ms / 2320 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 62 ms / 2402 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 62 ms / 2402 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 62 ms / 2402 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 62 ms / 2320 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 62 ms / 2364 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 62 ms / 2364 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 63 ms / 2402 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 63 ms / 2320 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 63 ms / 2320 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 63 ms / 2364 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 63 ms / 2364 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 63 ms / 2364 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 63 ms / 2402 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 63 ms / 2320 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 63 ms / 2402 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 64 ms / 2364 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 64 ms / 2402 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 64 ms / 2320 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 64 ms / 2364 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 64 ms / 2320 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 65 ms / 2320 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 65 ms / 2402 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 65 ms / 2320 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 65 ms / 2320 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 66 ms / 2364 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 66 ms / 2364 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 67 ms / 2402 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 68 ms / 2364 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 69 ms / 2402 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 96 ms / 2332 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 97 ms / 2332 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 97 ms / 2332 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 97 ms / 2375 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 97 ms / 2332 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 97 ms / 2375 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 98 ms / 2332 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 98 ms / 2375 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 98 ms / 2332 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 98 ms / 2332 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 98 ms / 2375 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 98 ms / 2332 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 99 ms / 2332 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 99 ms / 2375 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 99 ms / 2375 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 99 ms / 2375 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 99 ms / 2375 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 99 ms / 2375 ms |
FP32 | cpu | Passed | Very fast | macos · aarch64 | 100 ms / 2332 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 100 ms / 2332 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 100 ms / 2410 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 100 ms / 2375 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 100 ms / 2332 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 100 ms / 2375 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 101 ms / 2332 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 101 ms / 2410 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 101 ms / 2410 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 101 ms / 2375 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 101 ms / 2410 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 101 ms / 2410 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 101 ms / 2410 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 102 ms / 2410 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 102 ms / 2410 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 102 ms / 2375 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 106 ms / 2410 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 106 ms / 2410 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 107 ms / 2410 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 110 ms / 2320 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 111 ms / 2364 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 112 ms / 2402 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 115 ms / 2410 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 118 ms / 2410 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 143 ms / 2332 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 145 ms / 2375 ms |
FP32 | cpu | Passed | Moderate | macos · aarch64 | 983 ms / 2410 ms |
Test setup & sample details
- Quality gate
- { "minimumTop1Accuracy": 1, "maximumCpuScoreDifference": 0.03, "finiteScoresInUnitInterval": true, "scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification", "performanceTargets": { "goodInferenceMs": 250, "acceptableInferenceMs": 1000, "description": "One image at pinned256px FixRes or256patch NaFlex default and four labels; good<=250ms,acceptable<=1000ms" }, "version": 2, "minimumLogOddsMargin": 1, "gateRevisionReason": "V1 required0.05absolute sigmoid margin, which conflated calibrated confidence with ranking despite the native independent sigmoid contract. All six semantic fixtures ranked correctly but dog/portrait failed absolute margin. Preserve fullv1failures;v2requires expected top1plus ≥1natural-log-odds separation (odds ratio≥e), finite scores and unchanged0.03CPU agreement. This measures ranking separation without treating sigmoid scores as calibrated probabilities." }
- Tested
- 2026-11-06T17:21:12.595510+00:00
- Environment
- { "hardware": "Apple M3 Max", "os": "macOS-27.2-arm64-arm-64bit-Mach-O", "runtime": "Transformers5.22.1/Torch2.10.0/Pillow12.0.0", "interpreterScope": "Existing isolated private probe venv from app-ownedCPython3.13.18; managed packaged provision is separate evidence" }
- Runtime
- transformers 5.22.1
- Artifact revision
- 94dffa8cb1179de3e03f091dbc3917e5d5a9ae84
- Fixture
{ "file": "model-samples/desktop/siglip2-family/siglip2-base-patch32-256-dog-input.json", "size": 567, "sha1": "f9b27f8c462739815e725c2824899b4eb249c8be", "sha256": "d280bd6cb58649d1f878f8ec5d188b9ff0b5feb0fd2127e374996c7fa9f3b7c8", "url": "https://assets.localai.tools/samples/v1/files/d280bd6cb58649d1f878f8ec5d188b9ff0b5feb0fd2127e374996c7fa9f3b7c8/siglip2-base-patch32-256-dog-input.json" }- Rerun when
- Original weights/processor/tokenizer artifact revision; Pinned runtime or relevant OS/driver/backend behavior; Fixture bytes/labels/expected result or quality gate; Image preprocessing/prompt template/sigmoid scoring
Per-case quality & failure details
[
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2410.0790830016194,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 1002.7598750020843,
"inferenceMs": 983.0668750000768,
"imagesPerSecond": 1.0172247945999826,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2410.0790830016194,
"totalMs": 3412.8389580037037,
"initializationReused": false,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1456504832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005093280871733441,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.687606808947549,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2410.0790830016194,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 112.93325000224286,
"inferenceMs": 101.44029199727811,
"imagesPerSecond": 9.858015787521909,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2410.0790830016194,
"totalMs": 2523.0123330038623,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456504832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0003477198198424958,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.562642080253941,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2410.0790830016194,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 113.06358299771091,
"inferenceMs": 100.06179199990584,
"imagesPerSecond": 9.993824615902753,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2410.0790830016194,
"totalMs": 2523.1426659993303,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456504832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0006585183471088385,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.479529353697282,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2410.0790830016194,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 113.06245800005854,
"inferenceMs": 100.64599999896018,
"imagesPerSecond": 9.935814637544775,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2410.0790830016194,
"totalMs": 2523.141541001678,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456504832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.05175861990892372,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.049413684490457,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2410.0790830016194,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 109.16208299749997,
"inferenceMs": 101.04437499830965,
"imagesPerSecond": 9.896641945845365,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2410.0790830016194,
"totalMs": 2519.2411659991194,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456504832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.060531910759891616,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.872981076805981,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2410.0790830016194,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 109.77387500315672,
"inferenceMs": 101.80741599833709,
"imagesPerSecond": 9.822467157169905,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2410.0790830016194,
"totalMs": 2519.852958004776,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456504832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0007585791172459722,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.2739829259827076,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2410.0790830016194,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 104.13199999675271,
"inferenceMs": 101.68966700075543,
"imagesPerSecond": 9.8338408364792,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2410.0790830016194,
"totalMs": 2514.211082998372,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456504832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01382752540121146,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.530035044484616,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2410.0790830016194,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 104.99408300165669,
"inferenceMs": 100.67574999993667,
"imagesPerSecond": 9.932878573048912,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2410.0790830016194,
"totalMs": 2515.073166003276,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456504832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0007305046892724931,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 1.770114967177289,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2410.0790830016194,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 110.54449999937788,
"inferenceMs": 107.33970799992676,
"imagesPerSecond": 9.316216884069428,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2410.0790830016194,
"totalMs": 2520.6235830009973,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456504832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2410.0790830016194,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 117.55954100226518,
"inferenceMs": 106.11316699942108,
"imagesPerSecond": 9.423901182833,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2527.6386240038846,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456504832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2410.0790830016194,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 117.209375002858,
"inferenceMs": 105.64916599832941,
"imagesPerSecond": 9.465290052699636,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2527.2884580044774,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456504832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2410.0790830016194,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 131.47312500223052,
"inferenceMs": 118.43729099928169,
"imagesPerSecond": 8.443286667254698,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2541.55220800385,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456504832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2410.0790830016194,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 126.88775000060559,
"inferenceMs": 114.61041599977762,
"imagesPerSecond": 8.725210455583202,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2536.966833002225,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456504832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2410.0790830016194,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 112.70400000285008,
"inferenceMs": 101.26504199797637,
"imagesPerSecond": 9.875076139503141,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2522.7830830044695,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1456504832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2374.589458999253,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 163.61979199791676,
"inferenceMs": 145.10520799740334,
"imagesPerSecond": 6.891551404673876,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2374.589458999253,
"totalMs": 2538.20925099717,
"initializationReused": false,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454096384
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005093280871733441,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.687606808947549,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2374.589458999253,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 109.60941699886462,
"inferenceMs": 98.1644999992568,
"imagesPerSecond": 10.186982055708233,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2374.589458999253,
"totalMs": 2484.1988759981177,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454096384
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0003477198198424958,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.562642080253941,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2374.589458999253,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 112.16254200189724,
"inferenceMs": 99.34775000147056,
"imagesPerSecond": 10.06565322299899,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2374.589458999253,
"totalMs": 2486.7520010011503,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454096384
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0006585183471088385,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.479529353697282,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2374.589458999253,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 112.65962499965099,
"inferenceMs": 100.2362089966482,
"imagesPerSecond": 9.976434763543772,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2374.589458999253,
"totalMs": 2487.249083998904,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454096384
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.05175861990892372,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.049413684490457,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2374.589458999253,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 105.42899999927613,
"inferenceMs": 97.36762499960605,
"imagesPerSecond": 10.270354237397143,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2374.589458999253,
"totalMs": 2480.0184589985292,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454096384
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.060531910759891616,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.872981076805981,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2374.589458999253,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 107.15283299941802,
"inferenceMs": 99.04933299912955,
"imagesPerSecond": 10.095979142118889,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2374.589458999253,
"totalMs": 2481.742291998671,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454096384
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0007585791172459722,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.2739829259827076,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2374.589458999253,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 99.4872499977646,
"inferenceMs": 97.20999999990454,
"imagesPerSecond": 10.287007509525584,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2374.589458999253,
"totalMs": 2474.0767089970177,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454096384
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01382752540121146,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.530035044484616,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2374.589458999253,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 101.98362500159419,
"inferenceMs": 97.61762500056648,
"imagesPerSecond": 10.24405172727975,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2374.589458999253,
"totalMs": 2476.5730840008473,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454096384
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0007305046892724931,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 1.770114967177289,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2374.589458999253,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 102.39287500007777,
"inferenceMs": 99.35083300297265,
"imagesPerSecond": 10.065340871073314,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2374.589458999253,
"totalMs": 2476.982333999331,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454096384
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2374.589458999253,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 110.00220799905946,
"inferenceMs": 98.97999999884632,
"imagesPerSecond": 10.103051121556431,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2484.5916669983126,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454096384
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2374.589458999253,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 111.97887500020443,
"inferenceMs": 100.97483299978194,
"imagesPerSecond": 9.90345782500239,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2486.5683339994575,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454096384
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2374.589458999253,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 111.20058300002711,
"inferenceMs": 100.42949999842676,
"imagesPerSecond": 9.957233681494632,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2485.79004199928,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454096384
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2374.589458999253,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 110.56104200179107,
"inferenceMs": 99.4817079990753,
"imagesPerSecond": 10.052099226214484,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2485.150501001044,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454096384
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2374.589458999253,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 113.0113330000313,
"inferenceMs": 102.05791699991096,
"imagesPerSecond": 9.798357926518062,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2487.6007919992844,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1454096384
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2331.86058299907,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 161.8551669998851,
"inferenceMs": 143.41108400185476,
"imagesPerSecond": 6.97296172719165,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2331.86058299907,
"totalMs": 2493.7157499989553,
"initializationReused": false,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1455423488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005093280871733441,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.687606808947549,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2331.86058299907,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 110.73537500124075,
"inferenceMs": 99.58929200001876,
"imagesPerSecond": 10.04124017670305,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2331.86058299907,
"totalMs": 2442.595958000311,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1455423488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0003477198198424958,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.562642080253941,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2331.86058299907,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 109.84204199849046,
"inferenceMs": 97.15929099911591,
"imagesPerSecond": 10.292376464635785,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2331.86058299907,
"totalMs": 2441.7026249975606,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1455423488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0006585183471088385,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.479529353697282,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2331.86058299907,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 110.98012499860488,
"inferenceMs": 98.32154100149637,
"imagesPerSecond": 10.17071121764437,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2331.86058299907,
"totalMs": 2442.840707997675,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1455423488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.05175861990892372,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.049413684490457,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2331.86058299907,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 104.4419169993489,
"inferenceMs": 96.4692079978704,
"imagesPerSecond": 10.36600196844236,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2331.86058299907,
"totalMs": 2436.302499998419,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1455423488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.060531910759891616,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.872981076805981,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2331.86058299907,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 106.91470799793024,
"inferenceMs": 98.97716699924786,
"imagesPerSecond": 10.103340298753945,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2331.86058299907,
"totalMs": 2438.7752909970004,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1455423488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0007585791172459722,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.2739829259827076,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2331.86058299907,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 100.14083399801166,
"inferenceMs": 97.72145899842144,
"imagesPerSecond": 10.233166903659857,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2331.86058299907,
"totalMs": 2432.001416997082,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1455423488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01382752540121146,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.530035044484616,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2331.86058299907,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 101.66233299969463,
"inferenceMs": 97.78720800022711,
"imagesPerSecond": 10.226286448404146,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2331.86058299907,
"totalMs": 2433.522915998765,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1455423488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0007305046892724931,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 1.770114967177289,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2331.86058299907,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 99.88162500303588,
"inferenceMs": 96.85945899764192,
"imagesPerSecond": 10.32423689279893,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2331.86058299907,
"totalMs": 2431.742208002106,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1455423488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2331.86058299907,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 108.49970799972652,
"inferenceMs": 97.5083330013149,
"imagesPerSecond": 10.25553374998745,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2440.3602909987967,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1455423488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2331.86058299907,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 108.33570899922051,
"inferenceMs": 97.21016700132168,
"imagesPerSecond": 10.286989837044555,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2440.1962919982907,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1455423488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2331.86058299907,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 112.10987500089686,
"inferenceMs": 100.55437499977415,
"imagesPerSecond": 9.94486813728638,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2443.970457999967,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1455423488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2331.86058299907,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 111.21316700155148,
"inferenceMs": 100.0473329986562,
"imagesPerSecond": 9.995268939487188,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2443.0737500006217,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1455423488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2331.86058299907,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 111.40466699725948,
"inferenceMs": 100.31949999756762,
"imagesPerSecond": 9.968151755384012,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2443.2652499963297,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 1,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1455423488
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2402.0327500002168,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 130.67429199873004,
"inferenceMs": 112.36229199857917,
"imagesPerSecond": 8.899782856090592,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2402.0327500002168,
"totalMs": 2532.707041998947,
"initializationReused": false,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458913280
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005093280871733441,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.687606808947549,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2402.0327500002168,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 73.18791699799476,
"inferenceMs": 62.073959001281764,
"imagesPerSecond": 16.10981506720638,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2402.0327500002168,
"totalMs": 2475.2206669982115,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458913280
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0003477198198424958,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.562642080253941,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2402.0327500002168,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 76.37870799953816,
"inferenceMs": 63.437500000873115,
"imagesPerSecond": 15.763546797812596,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2402.0327500002168,
"totalMs": 2478.411457999755,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458913280
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0006585183471088385,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.479529353697282,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2402.0327500002168,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 79.03845799955889,
"inferenceMs": 66.66687500182888,
"imagesPerSecond": 14.999953124734988,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2402.0327500002168,
"totalMs": 2481.0712079997757,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458913280
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.05175861990892372,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.049413684490457,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2402.0327500002168,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 68.8314580002043,
"inferenceMs": 60.77270899913856,
"imagesPerSecond": 16.454754386778692,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2402.0327500002168,
"totalMs": 2470.864208000421,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458913280
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.060531910759891616,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.872981076805981,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2402.0327500002168,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 69.78041699767346,
"inferenceMs": 61.90525000056368,
"imagesPerSecond": 16.15371878783939,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2402.0327500002168,
"totalMs": 2471.81316699789,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458913280
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0007585791172459722,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.2739829259827076,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2402.0327500002168,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 63.38087499898393,
"inferenceMs": 61.14037500083214,
"imagesPerSecond": 16.35580416355624,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2402.0327500002168,
"totalMs": 2465.4136249992007,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458913280
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01382752540121146,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.530035044484616,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2402.0327500002168,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 67.0637919974979,
"inferenceMs": 63.054250000277534,
"imagesPerSecond": 15.859359202521613,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2402.0327500002168,
"totalMs": 2469.0965419977147,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458913280
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0007305046892724931,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 1.770114967177289,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2402.0327500002168,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 65.98154099992826,
"inferenceMs": 62.608457999886014,
"imagesPerSecond": 15.972282850374954,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2402.0327500002168,
"totalMs": 2468.014291000145,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458913280
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2402.0327500002168,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 80.25841600101558,
"inferenceMs": 64.9884580016078,
"imagesPerSecond": 15.387347703730104,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2482.2911660012323,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458913280
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2402.0327500002168,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 74.27637499858974,
"inferenceMs": 63.71125000077882,
"imagesPerSecond": 15.695815103106215,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2476.3091249988065,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458913280
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2402.0327500002168,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 73.02949999939301,
"inferenceMs": 62.09087499883026,
"imagesPerSecond": 16.105426119680857,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2475.06224999961,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458913280
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2402.0327500002168,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 72.32091600235435,
"inferenceMs": 61.376624998956686,
"imagesPerSecond": 16.292847643822036,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2474.353666002571,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458913280
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2402.0327500002168,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 79.92137500332319,
"inferenceMs": 68.89929099997971,
"imagesPerSecond": 14.513937451116798,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2481.95412500354,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458913280
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2319.563291999657,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 128.96020900006988,
"inferenceMs": 110.16199999721721,
"imagesPerSecond": 9.07754034989616,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2319.563291999657,
"totalMs": 2448.523500999727,
"initializationReused": false,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458552832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005093280871733441,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.687606808947549,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2319.563291999657,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 75.05020900134696,
"inferenceMs": 63.96720799966715,
"imagesPerSecond": 15.633009963561383,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2319.563291999657,
"totalMs": 2394.613501001004,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458552832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0003477198198424958,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.562642080253941,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2319.563291999657,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 78.46649999919464,
"inferenceMs": 65.20570800057612,
"imagesPerSecond": 15.336080700038783,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2319.563291999657,
"totalMs": 2398.0297919988516,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458552832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0006585183471088385,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.479529353697282,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2319.563291999657,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 76.78383399979793,
"inferenceMs": 64.29874999957974,
"imagesPerSecond": 15.552401874166076,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2319.563291999657,
"totalMs": 2396.347125999455,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458552832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.05175861990892372,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.049413684490457,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2319.563291999657,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 69.03720799891744,
"inferenceMs": 61.18074999903911,
"imagesPerSecond": 16.345010481494683,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2319.563291999657,
"totalMs": 2388.6004999985744,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458552832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.060531910759891616,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.872981076805981,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2319.563291999657,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 72.8850000014063,
"inferenceMs": 65.01291599852266,
"imagesPerSecond": 15.381558950881757,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2319.563291999657,
"totalMs": 2392.4482920010632,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458552832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0007585791172459722,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.2739829259827076,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2319.563291999657,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 64.01858299796004,
"inferenceMs": 61.75341700145509,
"imagesPerSecond": 16.193435902930474,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2319.563291999657,
"totalMs": 2383.581874997617,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458552832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01382752540121146,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.530035044484616,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2319.563291999657,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 65.78629099749378,
"inferenceMs": 61.82037499820581,
"imagesPerSecond": 16.175896701193135,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2319.563291999657,
"totalMs": 2385.3495829971507,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458552832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0007305046892724931,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 1.770114967177289,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2319.563291999657,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 66.50879199878545,
"inferenceMs": 63.416791999770794,
"imagesPerSecond": 15.768694197013534,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2319.563291999657,
"totalMs": 2386.0720839984424,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458552832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2319.563291999657,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 75.54345799871953,
"inferenceMs": 64.50204100110568,
"imagesPerSecond": 15.50338538873302,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2395.1067499983765,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458552832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2319.563291999657,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 73.17054200029816,
"inferenceMs": 62.11812499896041,
"imagesPerSecond": 16.098360985891567,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2392.733833999955,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458552832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2319.563291999657,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 73.20999999865307,
"inferenceMs": 61.34770800053957,
"imagesPerSecond": 16.300527478405627,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2392.77329199831,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458552832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2319.563291999657,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 73.92554099715198,
"inferenceMs": 62.79545799770858,
"imagesPerSecond": 15.924718632301245,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2393.488832996809,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458552832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2319.563291999657,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 73.76558299802127,
"inferenceMs": 62.88029099960113,
"imagesPerSecond": 15.903234290158474,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2393.328874997678,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1458552832
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2363.8917919997766,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 129.00312499914435,
"inferenceMs": 110.75849999906495,
"imagesPerSecond": 9.02865242855801,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2363.8917919997766,
"totalMs": 2492.894916998921,
"initializationReused": false,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1457979392
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005093280871733441,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.687606808947549,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2363.8917919997766,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 74.20762499896227,
"inferenceMs": 62.936042002547765,
"imagesPerSecond": 15.889146634920547,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2363.8917919997766,
"totalMs": 2438.099416998739,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1457979392
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0003477198198424958,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.562642080253941,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2363.8917919997766,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 77.11054200262879,
"inferenceMs": 64.09649999841349,
"imagesPerSecond": 15.60147590000627,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2363.8917919997766,
"totalMs": 2441.0023340024054,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1457979392
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0006585183471088385,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.479529353697282,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2363.8917919997766,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 74.3268749974959,
"inferenceMs": 61.806040997907985,
"imagesPerSecond": 16.179648200308577,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2363.8917919997766,
"totalMs": 2438.2186669972725,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1457979392
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.05175861990892372,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.049413684490457,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2363.8917919997766,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 70.31125000139582,
"inferenceMs": 62.471833000017796,
"imagesPerSecond": 16.007214003144668,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2363.8917919997766,
"totalMs": 2434.2030420011724,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1457979392
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.060531910759891616,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.872981076805981,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2363.8917919997766,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 71.31495900102891,
"inferenceMs": 63.043458001629915,
"imagesPerSecond": 15.862074062849569,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2363.8917919997766,
"totalMs": 2435.2067510008055,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1457979392
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0007585791172459722,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.2739829259827076,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2363.8917919997766,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 68.80491599804373,
"inferenceMs": 66.22616700042272,
"imagesPerSecond": 15.09977166568642,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2363.8917919997766,
"totalMs": 2432.6967079978203,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1457979392
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.01382752540121146,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.530035044484616,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2363.8917919997766,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 66.96545899831108,
"inferenceMs": 63.00341599853709,
"imagesPerSecond": 15.872155249855332,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2363.8917919997766,
"totalMs": 2430.8572509980877,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1457979392
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0007305046892724931,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 1.770114967177289,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2363.8917919997766,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 64.37416700282483,
"inferenceMs": 61.332041997957276,
"imagesPerSecond": 16.304691111267843,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"processInitMs": 2363.8917919997766,
"totalMs": 2428.2659590026014,
"initializationReused": true,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1457979392
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2363.8917919997766,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 72.58858299974236,
"inferenceMs": 61.60699999963981,
"imagesPerSecond": 16.231921697304635,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2436.480374999519,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1457979392
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2363.8917919997766,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 73.18216700150515,
"inferenceMs": 62.16916699850117,
"imagesPerSecond": 16.085143943204336,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2437.0739590012818,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1457979392
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2363.8917919997766,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 79.17541600181721,
"inferenceMs": 68.03808300173841,
"imagesPerSecond": 14.697651019568694,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2443.067208001594,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1457979392
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2363.8917919997766,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 74.24587499917834,
"inferenceMs": 63.55925000025309,
"imagesPerSecond": 15.733351164402004,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2438.137666998955,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1457979392
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748924737417838,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.8560170419110085,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2363.8917919997766,
"placement": {
"cpu/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 77.42599999983213,
"inferenceMs": 66.07349999831058,
"imagesPerSecond": 15.134660643458705,
"outputPlacement": "cpu",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": null,
"totalMs": 2441.3177919996087,
"device": "cpu",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1457979392
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 258.9240000015707,
"inferenceMs": 225.76904199740966,
"imagesPerSecond": 4.429305236682864,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573584896,
"processInitMs": 2922.090541000216,
"totalMs": 3181.0145410017867,
"initializationReused": false,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 2.9336661100387573e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005093310257507255,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.687615400971553,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 79.12487499925192,
"inferenceMs": 66.4443339992431,
"imagesPerSecond": 15.050192240791992,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"processInitMs": 2922.090541000216,
"totalMs": 3001.215415999468,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 6.693881005048752e-10,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0003477191249885436,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.562621163013367,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 78.99116599946865,
"inferenceMs": 64.90470800054027,
"imagesPerSecond": 15.407202817886123,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"processInitMs": 2922.090541000216,
"totalMs": 3001.0817069996847,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.280568540096283e-9,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0006585169826394122,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.479497019661535,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 86.55562499916414,
"inferenceMs": 72.73887499832199,
"imagesPerSecond": 13.747806795514352,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"processInitMs": 2922.090541000216,
"totalMs": 3008.64616599938,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 6.556510925292969e-7,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.05175927549271364,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.049416545730725,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 79.89799999995739,
"inferenceMs": 69.6641249996901,
"imagesPerSecond": 14.354590687882013,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"processInitMs": 2922.090541000216,
"totalMs": 3001.9885410001734,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000020042061805725098,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.06052990672560554,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.872964874525058,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 74.46833299763966,
"inferenceMs": 65.53033300224342,
"imagesPerSecond": 15.260108627340031,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"processInitMs": 2922.090541000216,
"totalMs": 2996.5588739978557,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.032657206058502e-9,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.000758588284952566,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.273994409008412,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 64.32904199755285,
"inferenceMs": 60.63712499963003,
"imagesPerSecond": 16.49154705151508,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"processInitMs": 2922.090541000216,
"totalMs": 2986.419582997769,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 7.543712854385376e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.013827600761942449,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.530036839311882,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 65.9516670020821,
"inferenceMs": 60.594999999011634,
"imagesPerSecond": 16.503011799922618,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"processInitMs": 2922.090541000216,
"totalMs": 2988.042208002298,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 4.292814992368221e-9,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0007305011531570926,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 1.770087221003772,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 61.0966250023921,
"inferenceMs": 55.61391700030072,
"imagesPerSecond": 17.98111073518869,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"processInitMs": 2922.090541000216,
"totalMs": 2983.187166002608,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 80.9422500024084,
"inferenceMs": 68.13941600194084,
"imagesPerSecond": 14.675793522672935,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3003.0327910026244,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 83.93250000153785,
"inferenceMs": 69.53633299781359,
"imagesPerSecond": 14.380971168431369,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3006.023041001754,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 85.48333400176489,
"inferenceMs": 70.37479100108612,
"imagesPerSecond": 14.20963367386152,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3007.573875001981,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 83.16670799831627,
"inferenceMs": 68.67358300223714,
"imagesPerSecond": 14.561640099183752,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3005.2572489985323,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 84.10983400244731,
"inferenceMs": 71.41029199920013,
"imagesPerSecond": 14.003583685264879,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3006.2003750026633,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 78.21975000115344,
"inferenceMs": 65.7523330010008,
"imagesPerSecond": 15.208585830479645,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3000.3102910013695,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 83.81658399957814,
"inferenceMs": 71.09508400026243,
"imagesPerSecond": 14.065670138265942,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3005.907124999794,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 81.71166600004653,
"inferenceMs": 69.35283399798209,
"imagesPerSecond": 14.41902143507353,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3003.8022070002626,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 85.72354199714027,
"inferenceMs": 71.0964169993531,
"imagesPerSecond": 14.065406418569573,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3007.8140829973563,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 77.99312499992084,
"inferenceMs": 65.60670799808577,
"imagesPerSecond": 15.242343816872772,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3000.083666000137,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 83.15766699888627,
"inferenceMs": 70.2129999990575,
"imagesPerSecond": 14.242376768026197,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3005.2482079991023,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 81.81754200268188,
"inferenceMs": 68.76091699814424,
"imagesPerSecond": 14.543145200157651,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3003.908083002898,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 83.8968339994608,
"inferenceMs": 69.45058299970697,
"imagesPerSecond": 14.398727221688251,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3005.987374999677,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 80.2094999999099,
"inferenceMs": 67.89412500074832,
"imagesPerSecond": 14.728814900979698,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3002.300041000126,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 80.36991599874455,
"inferenceMs": 67.87650000114809,
"imagesPerSecond": 14.732639425767175,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3002.4604569989606,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 79.79308300127741,
"inferenceMs": 64.94791699878988,
"imagesPerSecond": 15.396952607712302,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3001.8836240014934,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 84.06399999876157,
"inferenceMs": 70.8384580029815,
"imagesPerSecond": 14.116625745268358,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3006.1545409989776,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 82.47937499982072,
"inferenceMs": 69.95704200016917,
"imagesPerSecond": 14.29448660790406,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3004.5699160000368,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 78.98754200141411,
"inferenceMs": 66.55079199845204,
"imagesPerSecond": 15.026117195168164,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3001.07808300163,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 83.20991700020386,
"inferenceMs": 69.37800000014249,
"imagesPerSecond": 14.413791115309552,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3005.30045800042,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 78.4760839997034,
"inferenceMs": 65.82820900075603,
"imagesPerSecond": 15.191055858568399,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3000.5666249999194,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 83.88249999916297,
"inferenceMs": 71.07095899846172,
"imagesPerSecond": 14.070444723021739,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3005.973040999379,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 83.26904200293939,
"inferenceMs": 69.49291600176366,
"imagesPerSecond": 14.389955948526048,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3005.3595830031554,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 77.41650000025402,
"inferenceMs": 65.010207999876,
"imagesPerSecond": 15.382199669348964,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 2999.50704100047,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 82.19850000023143,
"inferenceMs": 69.7815419989638,
"imagesPerSecond": 14.330437123542628,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3004.2890410004475,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 84.12299999690731,
"inferenceMs": 69.5656659991073,
"imagesPerSecond": 14.374907300001016,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 3006.2135409971233,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 59.18958300026134,
"inferenceMs": 46.5896249988873,
"imagesPerSecond": 21.464006203610417,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 2981.2801240004774,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 45.10933300116449,
"inferenceMs": 32.302540999808116,
"imagesPerSecond": 30.957316949336594,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 2967.1998740013805,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 50.67529199732235,
"inferenceMs": 38.43808400051785,
"imagesPerSecond": 26.015864890313672,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 2972.7658329975384,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2922.090541000216,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 47.8705829991668,
"inferenceMs": 35.14620799978729,
"imagesPerSecond": 28.452571611880636,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 2969.961123999383,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 940736512
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2683.46158300119,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 159.67545899911784,
"inferenceMs": 126.41375000021071,
"imagesPerSecond": 7.910531884374391,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573584896,
"processInitMs": 2683.46158300119,
"totalMs": 2843.1370420003077,
"initializationReused": false,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 798081024
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 2.9336661100387573e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005093310257507255,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.687615400971553,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2683.46158300119,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 52.07133300064015,
"inferenceMs": 38.57191700080875,
"imagesPerSecond": 25.925597630499738,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573584896,
"processInitMs": 2683.46158300119,
"totalMs": 2735.53291600183,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 798081024
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 6.693881005048752e-10,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0003477191249885436,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.562621163013367,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2683.46158300119,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 49.576334000448696,
"inferenceMs": 34.509083001466934,
"imagesPerSecond": 28.977878083792938,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573584896,
"processInitMs": 2683.46158300119,
"totalMs": 2733.0379170016386,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 798081024
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.280568540096283e-9,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0006585169826394122,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.479497019661535,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2683.46158300119,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 49.55162500118604,
"inferenceMs": 33.219749999261694,
"imagesPerSecond": 30.10257452335508,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573584896,
"processInitMs": 2683.46158300119,
"totalMs": 2733.013208002376,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 798081024
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 6.556510925292969e-7,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.05175927549271364,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.049416545730725,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2683.46158300119,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 42.13683300258708,
"inferenceMs": 32.68495799784432,
"imagesPerSecond": 30.59511350958301,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573584896,
"processInitMs": 2683.46158300119,
"totalMs": 2725.598416003777,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 798081024
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000020042061805725098,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.06052990672560554,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.872964874525058,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2683.46158300119,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.20941699715331,
"inferenceMs": 41.669209000247065,
"imagesPerSecond": 23.99853570520311,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573584896,
"processInitMs": 2683.46158300119,
"totalMs": 2734.670999998343,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 798081024
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.032657206058502e-9,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.000758588284952566,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.273994409008412,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2683.46158300119,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 23.6696669999219,
"inferenceMs": 19.90562500213855,
"imagesPerSecond": 50.23705610311485,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573584896,
"processInitMs": 2683.46158300119,
"totalMs": 2707.1312500011118,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 798081024
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 7.543712854385376e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.013827600761942449,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.530036839311882,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2683.46158300119,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 32.88583399989875,
"inferenceMs": 27.738000000681495,
"imagesPerSecond": 36.05162592744362,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"processInitMs": 2683.46158300119,
"totalMs": 2716.3474170010886,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 798081024
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 4.292814992368221e-9,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0007305011531570926,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 1.770087221003772,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2683.46158300119,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 24.800375002087094,
"inferenceMs": 20.372999999381136,
"imagesPerSecond": 49.08457272028551,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"processInitMs": 2683.46158300119,
"totalMs": 2708.261958003277,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 798081024
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2683.46158300119,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 49.935125000047265,
"inferenceMs": 37.31849999894621,
"imagesPerSecond": 26.796361054925512,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 2733.396708001237,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 798081024
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2683.46158300119,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 47.40641699754633,
"inferenceMs": 32.93458300322527,
"imagesPerSecond": 30.363220323817988,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 2730.867999998736,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 798081024
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2683.46158300119,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.486291998415254,
"inferenceMs": 35.88062500057276,
"imagesPerSecond": 27.87019456835094,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 2731.947874999605,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 798081024
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2683.46158300119,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 44.49929100155714,
"inferenceMs": 32.1636659973592,
"imagesPerSecond": 31.090983225671632,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 2727.960874002747,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 798081024
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2683.46158300119,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 52.52699999982724,
"inferenceMs": 38.96499999973457,
"imagesPerSecond": 25.664057487663595,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 2735.988583001017,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 798081024
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2541.494249999232,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 166.63825000068755,
"inferenceMs": 137.78829199873144,
"imagesPerSecond": 7.257510674486092,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573584896,
"processInitMs": 2541.494249999232,
"totalMs": 2708.1324999999197,
"initializationReused": false,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 797425664
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 2.9336661100387573e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005093310257507255,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.687615400971553,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2541.494249999232,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 73.39191700157244,
"inferenceMs": 60.995709001872456,
"imagesPerSecond": 16.394595888200953,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573584896,
"processInitMs": 2541.494249999232,
"totalMs": 2614.8861670008046,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 797425664
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 6.693881005048752e-10,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0003477191249885436,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.562621163013367,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2541.494249999232,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.06337499880465,
"inferenceMs": 39.401417001499794,
"imagesPerSecond": 25.379797888028634,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573584896,
"processInitMs": 2541.494249999232,
"totalMs": 2595.5576249980368,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797425664
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 1.280568540096283e-9,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0006585169826394122,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.479497019661535,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2541.494249999232,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.046542001131456,
"inferenceMs": 37.052874999062624,
"imagesPerSecond": 26.988459060877148,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573584896,
"processInitMs": 2541.494249999232,
"totalMs": 2592.5407920003636,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797425664
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 6.556510925292969e-7,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.05175927549271364,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.049416545730725,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2541.494249999232,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 45.42333300196333,
"inferenceMs": 36.39191600086633,
"imagesPerSecond": 27.478630143469072,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573584896,
"processInitMs": 2541.494249999232,
"totalMs": 2586.9175830011955,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797425664
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0000020042061805725098,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.06052990672560554,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.872964874525058,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2541.494249999232,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 45.70233400227153,
"inferenceMs": 36.9738330009568,
"imagesPerSecond": 27.04615450538012,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573584896,
"processInitMs": 2541.494249999232,
"totalMs": 2587.1965840015037,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797425664
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 8.032657206058502e-9,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.000758588284952566,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.273994409008412,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2541.494249999232,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 22.644083001068793,
"inferenceMs": 19.142459001159295,
"imagesPerSecond": 52.23989247877916,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573584896,
"processInitMs": 2541.494249999232,
"totalMs": 2564.138333000301,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797425664
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 7.543712854385376e-8,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.013827600761942449,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.530036839311882,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2541.494249999232,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 24.00491699881968,
"inferenceMs": 18.63062499978696,
"imagesPerSecond": 53.67506457842584,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"processInitMs": 2541.494249999232,
"totalMs": 2565.499166998052,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797425664
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 4.292814992368221e-9,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0007305011531570926,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 1.770087221003772,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2541.494249999232,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 23.160958000516985,
"inferenceMs": 19.26308399924892,
"imagesPerSecond": 51.91276744881508,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"processInitMs": 2541.494249999232,
"totalMs": 2564.655207999749,
"initializationReused": true,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797425664
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2541.494249999232,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 49.114916000689846,
"inferenceMs": 37.07087500151829,
"imagesPerSecond": 26.97535464051074,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 2590.609165999922,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797425664
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2541.494249999232,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 45.294958999875234,
"inferenceMs": 32.507915999303805,
"imagesPerSecond": 30.761738156989708,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 2586.7892089991074,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797425664
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2541.494249999232,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 69.27795800220338,
"inferenceMs": 56.967708998854505,
"imagesPerSecond": 17.553804033441256,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 2610.7722080014355,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 797425664
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2541.494249999232,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 44.337583000015,
"inferenceMs": 32.344999999622814,
"imagesPerSecond": 30.916679548977008,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 2585.831832999247,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797425664
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 5.587935447692871e-9,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005748930499976268,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.856028536443957,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2541.494249999232,
"placement": {
"mps:0/torch.float32": 408
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.15837500063935,
"inferenceMs": 35.2033749986731,
"imagesPerSecond": 28.40636728829814,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1573568512,
"totalMs": 2589.6526249998715,
"device": "mps",
"precision": "fp32",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 797425664
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00014554383233189583,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005894367104701814,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.875000026560944,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2487.790250001126,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 418.139291999978,
"inferenceMs": 389.71658399896114,
"imagesPerSecond": 2.5659672722643636,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2487.790250001126,
"totalMs": 2905.929542001104,
"initializationReused": false,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1177092096
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00010944250971078873,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005202622503929888,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.703124971194826,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2487.790250001126,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 45.43537500285311,
"inferenceMs": 31.905874999210937,
"imagesPerSecond": 31.342190114664806,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2487.790250001126,
"totalMs": 2533.225625003979,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1177092096
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00000791042111814022,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.000355620914319843,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.578124937789591,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2487.790250001126,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 52.31104100312223,
"inferenceMs": 37.76649999781512,
"imagesPerSecond": 26.478492845719153,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2487.790250001126,
"totalMs": 2540.101291004248,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1177092096
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000021223328076303005,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0006796888617373043,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.492187516301669,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2487.790250001126,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 49.11675000039395,
"inferenceMs": 35.21845799696166,
"imagesPerSecond": 28.394201702024297,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2487.790250001126,
"totalMs": 2536.90700000152,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1177092096
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0020346082746982574,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.053793099950325995,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.070312478817474,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2487.790250001126,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 43.63374999957159,
"inferenceMs": 33.9217920009105,
"imagesPerSecond": 29.47957466318875,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2487.790250001126,
"totalMs": 2531.4240000006976,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1177092096
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002709578722715378,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.0632412958657369,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.898437517703979,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2487.790250001126,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 37.2209169981943,
"inferenceMs": 27.998249999654945,
"imagesPerSecond": 35.71651799710068,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"processInitMs": 2487.790250001126,
"totalMs": 2525.0111669993203,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1177092096
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000019829953089356422,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0007703329320065677,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.2656250957475468,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2487.790250001126,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 21.10279100088519,
"inferenceMs": 17.39862500107847,
"imagesPerSecond": 57.47580627423224,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"processInitMs": 2487.790250001126,
"totalMs": 2508.893041002011,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1177092096
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00032435450702905655,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014151743398542749,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.546875072090101,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2487.790250001126,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 23.03350000147475,
"inferenceMs": 17.408208001143066,
"imagesPerSecond": 57.44416656409077,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"processInitMs": 2487.790250001126,
"totalMs": 2510.8237500026007,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1177092096
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000030460301786661148,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0007562844402855262,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 1.7734373383248947,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2487.790250001126,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 22.589542000787333,
"inferenceMs": 18.007167000178015,
"imagesPerSecond": 55.533443988724834,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"processInitMs": 2487.790250001126,
"totalMs": 2510.3797920019133,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1177092096
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00014554383233189583,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005894367104701814,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.875000026560944,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2487.790250001126,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 46.81758299921057,
"inferenceMs": 34.65641699949629,
"imagesPerSecond": 28.85468512265808,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2534.6078330003365,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1177092096
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00014554383233189583,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005894367104701814,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.875000026560944,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2487.790250001126,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 52.726958001585444,
"inferenceMs": 40.57812500104774,
"imagesPerSecond": 24.64381979143146,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2540.5172080027114,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1177092096
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00014554383233189583,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005894367104701814,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.875000026560944,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2487.790250001126,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 45.94579200056614,
"inferenceMs": 34.17516699846601,
"imagesPerSecond": 29.261012829721828,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2533.736042001692,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1177092096
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00014554383233189583,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005894367104701814,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.875000026560944,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2487.790250001126,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 47.24487500061514,
"inferenceMs": 35.18279199852259,
"imagesPerSecond": 28.422985874514804,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2535.035125001741,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1177092096
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00014554383233189583,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005894367104701814,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.875000026560944,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2487.790250001126,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.602583000203595,
"inferenceMs": 38.92887500114739,
"imagesPerSecond": 25.687873075462008,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2539.3928330013296,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1177092096
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00014554383233189583,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005894367104701814,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.875000026560944,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2468.1856250026613,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 146.869541000342,
"inferenceMs": 116.80912500014529,
"imagesPerSecond": 8.560975009433177,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2468.1856250026613,
"totalMs": 2615.0551660030033,
"initializationReused": false,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1343602688
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00010944250971078873,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005202622503929888,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.703124971194826,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2468.1856250026613,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 47.963749999325955,
"inferenceMs": 35.3409580020525,
"imagesPerSecond": 28.29578077487098,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2468.1856250026613,
"totalMs": 2516.149375001987,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1343602688
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00000791042111814022,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.000355620914319843,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.578124937789591,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2468.1856250026613,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.31949999788776,
"inferenceMs": 34.087292002368486,
"imagesPerSecond": 29.336445967327563,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2468.1856250026613,
"totalMs": 2516.505125000549,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1343602688
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000021223328076303005,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0006796888617373043,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.492187516301669,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2468.1856250026613,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 47.94029199911165,
"inferenceMs": 33.799791999626905,
"imagesPerSecond": 29.5859808844693,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2468.1856250026613,
"totalMs": 2516.125917001773,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1343602688
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0020346082746982574,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.053793099950325995,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.070312478817474,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2468.1856250026613,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.64970900234766,
"inferenceMs": 38.09541699956753,
"imagesPerSecond": 26.24987672431443,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2468.1856250026613,
"totalMs": 2516.835334005009,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1343602688
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002709578722715378,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.0632412958657369,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.898437517703979,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2468.1856250026613,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 43.121457998495316,
"inferenceMs": 33.02145800262224,
"imagesPerSecond": 30.283338788995625,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2468.1856250026613,
"totalMs": 2511.3070830011566,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1343602688
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000019829953089356422,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0007703329320065677,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.2656250957475468,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2468.1856250026613,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 22.705167000822257,
"inferenceMs": 18.44612500281073,
"imagesPerSecond": 54.21192797119313,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2468.1856250026613,
"totalMs": 2490.8907920034835,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1343602688
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00032435450702905655,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014151743398542749,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.546875072090101,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2468.1856250026613,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 24.522334002540447,
"inferenceMs": 18.239917000755668,
"imagesPerSecond": 54.82481087817289,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2468.1856250026613,
"totalMs": 2492.7079590052017,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1343602688
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000030460301786661148,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0007562844402855262,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 1.7734373383248947,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2468.1856250026613,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 23.150375000113854,
"inferenceMs": 18.91737499681767,
"imagesPerSecond": 52.86145673848632,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2468.1856250026613,
"totalMs": 2491.336000002775,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1343602688
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00014554383233189583,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005894367104701814,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.875000026560944,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2468.1856250026613,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 49.22991700004786,
"inferenceMs": 36.450542000238784,
"imagesPerSecond": 27.43443430809476,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2517.415542002709,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1343602688
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00014554383233189583,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005894367104701814,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.875000026560944,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2468.1856250026613,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 51.357624997763196,
"inferenceMs": 38.96816699852934,
"imagesPerSecond": 25.661971732920875,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2519.5432500004245,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1343602688
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00014554383233189583,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005894367104701814,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.875000026560944,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2468.1856250026613,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 43.79437500028871,
"inferenceMs": 31.30270800102153,
"imagesPerSecond": 31.946117887543984,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2511.98000000295,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1343602688
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00014554383233189583,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005894367104701814,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.875000026560944,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2468.1856250026613,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 50.99216600137879,
"inferenceMs": 37.751208001282066,
"imagesPerSecond": 26.48921856927172,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2519.17779100404,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1343602688
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00014554383233189583,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005894367104701814,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.875000026560944,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2468.1856250026613,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 81.73529200212215,
"inferenceMs": 67.73779200011631,
"imagesPerSecond": 14.762807739559667,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2549.9209170047834,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1343602688
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00014554383233189583,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005894367104701814,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.875000026560944,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2567.566000001534,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 158.57395900093252,
"inferenceMs": 128.09850000121514,
"imagesPerSecond": 7.806492659871224,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2567.566000001534,
"totalMs": 2726.1399590024666,
"initializationReused": false,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1149714432
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00010944250971078873,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005202622503929888,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.703124971194826,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2567.566000001534,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 45.41241699917009,
"inferenceMs": 31.964875000994653,
"imagesPerSecond": 31.284339449751734,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2567.566000001534,
"totalMs": 2612.978417000704,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1149714432
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00000791042111814022,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.000355620914319843,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.578124937789591,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2567.566000001534,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 57.812457998807076,
"inferenceMs": 42.677417000959394,
"imagesPerSecond": 23.431596152539406,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2567.566000001534,
"totalMs": 2625.378458000341,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1149714432
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000021223328076303005,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0006796888617373043,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.492187516301669,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2567.566000001534,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 49.08454100223025,
"inferenceMs": 34.46583299955819,
"imagesPerSecond": 29.014241437681743,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2567.566000001534,
"totalMs": 2616.6505410037644,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1149714432
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0020346082746982574,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.053793099950325995,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.070312478817474,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2567.566000001534,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.345125000196276,
"inferenceMs": 37.79595900050481,
"imagesPerSecond": 26.45785492535442,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2567.566000001534,
"totalMs": 2615.9111250017304,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1149714432
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002709578722715378,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.0632412958657369,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.898437517703979,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2567.566000001534,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 43.09487499995157,
"inferenceMs": 32.79225000005681,
"imagesPerSecond": 30.495010253894367,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2567.566000001534,
"totalMs": 2610.6608750014857,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1149714432
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000019829953089356422,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0007703329320065677,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.2656250957475468,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2567.566000001534,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 23.270791996765183,
"inferenceMs": 19.25741599916364,
"imagesPerSecond": 51.928046838860965,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2567.566000001534,
"totalMs": 2590.8367919982993,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1149714432
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00032435450702905655,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.014151743398542749,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.546875072090101,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2567.566000001534,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 24.61220800250885,
"inferenceMs": 18.934583000373095,
"imagesPerSecond": 52.81341553602187,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2567.566000001534,
"totalMs": 2592.178208004043,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1149714432
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.000030460301786661148,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0007562844402855262,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 1.7734373383248947,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2567.566000001534,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 27.077209000708535,
"inferenceMs": 22.15983300266089,
"imagesPerSecond": 45.126693864521584,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2567.566000001534,
"totalMs": 2594.6432090022427,
"initializationReused": true,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1149714432
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00014554383233189583,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005894367104701814,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.875000026560944,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2567.566000001534,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 44.12520900223171,
"inferenceMs": 31.626207997760503,
"imagesPerSecond": 31.619345577908405,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2611.691209003766,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1149714432
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00014554383233189583,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005894367104701814,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.875000026560944,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2567.566000001534,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.01524999958929,
"inferenceMs": 40.341957999771694,
"imagesPerSecond": 24.78808787629146,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2620.5812500011234,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1149714432
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00014554383233189583,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005894367104701814,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.875000026560944,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2567.566000001534,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 46.97295900041354,
"inferenceMs": 33.72316700188094,
"imagesPerSecond": 29.653205463894423,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2614.5389590019477,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1149714432
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00014554383233189583,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005894367104701814,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.875000026560944,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2567.566000001534,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 46.525959001883166,
"inferenceMs": 34.564292000140995,
"imagesPerSecond": 28.931592175992517,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2614.0919590034173,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1149714432
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00014554383233189583,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005894367104701814,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.875000026560944,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2567.566000001534,
"placement": {
"mps:0/torch.float16": 408
},
"parameterDtype": "torch.float16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.16812499950174,
"inferenceMs": 40.389207999396604,
"imagesPerSecond": 24.759089111500764,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2620.734125001036,
"device": "mps",
"precision": "fp16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1149714432
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0005243769846856594,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.006270976529776817,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.812499953123904,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2671.21083299935,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 467.346292000002,
"inferenceMs": 436.794084002031,
"imagesPerSecond": 2.2894082970119856,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2671.21083299935,
"totalMs": 3138.557124999352,
"initializationReused": false,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "acceptable",
"observedResidentBytesAtCaseEnd": 1500037120
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0008003856055438519,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005890922835533274,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.687500004261356,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2671.21083299935,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 50.8294580031361,
"inferenceMs": 35.73545800099964,
"imagesPerSecond": 27.98341076171534,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2671.21083299935,
"totalMs": 2722.040291002486,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1500037120
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00005542128928937018,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0004029279872383995,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.562499950666638,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2671.21083299935,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.153166000702186,
"inferenceMs": 33.84225000263541,
"imagesPerSecond": 29.548862735844295,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2671.21083299935,
"totalMs": 2719.363999000052,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1500037120
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00009412708459421992,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0007523161411882029,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.500000024236441,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2671.21083299935,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 58.33037500269711,
"inferenceMs": 42.84358299992164,
"imagesPerSecond": 23.340718258830712,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2671.21083299935,
"totalMs": 2729.541208002047,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1500037120
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0016382932662963867,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.05339636312328366,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.000000004326502,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2671.21083299935,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 49.430999999458436,
"inferenceMs": 39.520665999589255,
"imagesPerSecond": 25.30321730940448,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"processInitMs": 2671.21083299935,
"totalMs": 2720.6418329988082,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1500037120
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.007005747407674789,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.06753655802913272,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.874999939047145,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2671.21083299935,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 46.963000000687316,
"inferenceMs": 35.05983300055959,
"imagesPerSecond": 28.522668661429133,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"processInitMs": 2671.21083299935,
"totalMs": 2718.173833000037,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1500037120
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00038677407428622246,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0005178638093639165,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.5000000226721921,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2671.21083299935,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 22.90829199773725,
"inferenceMs": 18.509833000280196,
"imagesPerSecond": 54.02533885556193,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"processInitMs": 2671.21083299935,
"totalMs": 2694.119124997087,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1500037120
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002058379352092743,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.015884946718870196,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.625000088039937,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2671.21083299935,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 28.110082999774022,
"inferenceMs": 21.654125001077773,
"imagesPerSecond": 46.18057760127587,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"processInitMs": 2671.21083299935,
"totalMs": 2699.320915999124,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1500037120
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00015164026990532875,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0008635782869532704,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 1.8125000765513555,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2671.21083299935,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 27.936500002397224,
"inferenceMs": 22.608667000895366,
"imagesPerSecond": 44.23082528308269,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"processInitMs": 2671.21083299935,
"totalMs": 2699.147333001747,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1500037120
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0005243769846856594,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.006270976529776817,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.812499953123904,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2671.21083299935,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.91670899814926,
"inferenceMs": 35.526499999832595,
"imagesPerSecond": 28.14800219567681,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2720.127541997499,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1500037120
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0005243769846856594,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.006270976529776817,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.812499953123904,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2671.21083299935,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.145458998391405,
"inferenceMs": 40.588625000964385,
"imagesPerSecond": 24.637444603660263,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2725.356291997741,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1500037120
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0005243769846856594,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.006270976529776817,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.812499953123904,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2671.21083299935,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 47.49900000024354,
"inferenceMs": 34.131415999581804,
"imagesPerSecond": 29.298520753204393,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2718.7098329995933,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1500037120
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0005243769846856594,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.006270976529776817,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.812499953123904,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2671.21083299935,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 48.26883400164661,
"inferenceMs": 34.12479200051166,
"imagesPerSecond": 29.304207919714386,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2719.4796670009964,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1500037120
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0005243769846856594,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.006270976529776817,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.812499953123904,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2671.21083299935,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 45.061500000883825,
"inferenceMs": 32.21958300127881,
"imagesPerSecond": 31.037024903776988,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2716.2723330002336,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1500037120
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0005243769846856594,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.006270976529776817,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.812499953123904,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2611.9960420001007,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 147.61704099873896,
"inferenceMs": 118.0759169983503,
"imagesPerSecond": 8.469127536091644,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2611.9960420001007,
"totalMs": 2759.6130829988397,
"initializationReused": false,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1417756672
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0008003856055438519,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005890922835533274,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.687500004261356,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2611.9960420001007,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 50.870041999587556,
"inferenceMs": 38.16779199769371,
"imagesPerSecond": 26.20010086149141,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2611.9960420001007,
"totalMs": 2662.8660839996883,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1417756672
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00005542128928937018,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0004029279872383995,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.562499950666638,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2611.9960420001007,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 45.22970800098847,
"inferenceMs": 29.806709000695264,
"imagesPerSecond": 33.54949383968134,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2611.9960420001007,
"totalMs": 2657.225750001089,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1417756672
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00009412708459421992,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0007523161411882029,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.500000024236441,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2611.9960420001007,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 45.48154199801502,
"inferenceMs": 31.76133399756509,
"imagesPerSecond": 31.484823656231278,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2611.9960420001007,
"totalMs": 2657.4775839981157,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1417756672
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0016382932662963867,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.05339636312328366,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.000000004326502,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2611.9960420001007,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 41.93887500150595,
"inferenceMs": 32.38858300028369,
"imagesPerSecond": 30.875077183563143,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2611.9960420001007,
"totalMs": 2653.9349170016067,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1417756672
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.007005747407674789,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.06753655802913272,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.874999939047145,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2611.9960420001007,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 36.9433330015454,
"inferenceMs": 27.394709002692252,
"imagesPerSecond": 36.50339924770596,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2611.9960420001007,
"totalMs": 2648.939375001646,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1417756672
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00038677407428622246,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0005178638093639165,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.5000000226721921,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2611.9960420001007,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 20.673582999734208,
"inferenceMs": 16.86849999896367,
"imagesPerSecond": 59.28209384719659,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2611.9960420001007,
"totalMs": 2632.669624999835,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1417756672
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002058379352092743,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.015884946718870196,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.625000088039937,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2611.9960420001007,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 22.89566599938553,
"inferenceMs": 17.196208002133062,
"imagesPerSecond": 58.15235544231364,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2611.9960420001007,
"totalMs": 2634.8917079994862,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1417756672
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00015164026990532875,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0008635782869532704,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 1.8125000765513555,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2611.9960420001007,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 27.643083998555085,
"inferenceMs": 22.50975000060862,
"imagesPerSecond": 44.42519352604813,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2611.9960420001007,
"totalMs": 2639.639125998656,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1417756672
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0005243769846856594,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.006270976529776817,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.812499953123904,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2611.9960420001007,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 78.14516600046773,
"inferenceMs": 65.83025000145426,
"imagesPerSecond": 15.190584875158592,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2690.1412080005684,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1417756672
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0005243769846856594,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.006270976529776817,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.812499953123904,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2611.9960420001007,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 80.54183399872272,
"inferenceMs": 62.74491599833709,
"imagesPerSecond": 15.937546239228412,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2692.5378759988234,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1417756672
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0005243769846856594,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.006270976529776817,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.812499953123904,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2611.9960420001007,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 76.46183399992879,
"inferenceMs": 63.7353330021142,
"imagesPerSecond": 15.689884290191571,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2688.4578760000295,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1417756672
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0005243769846856594,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.006270976529776817,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.812499953123904,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2611.9960420001007,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 82.74266699663713,
"inferenceMs": 67.4109169995063,
"imagesPerSecond": 14.834392476923638,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2694.738708996738,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1417756672
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0005243769846856594,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.006270976529776817,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.812499953123904,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2611.9960420001007,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 81.31720800156472,
"inferenceMs": 63.53812500310596,
"imagesPerSecond": 15.738582149711164,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2693.3132500016654,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1417756672
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0005243769846856594,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.006270976529776817,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.812499953123904,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2559.6974170002795,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 169.04900000008638,
"inferenceMs": 138.11737500145682,
"imagesPerSecond": 7.240218690729188,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2559.6974170002795,
"totalMs": 2728.746417000366,
"initializationReused": false,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1493696512
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0008003856055438519,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.005890922835533274,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.687500004261356,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2559.6974170002795,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 76.7610840011912,
"inferenceMs": 62.34237500029849,
"imagesPerSecond": 16.04045402497438,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2559.6974170002795,
"totalMs": 2636.4585010014707,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1493696512
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00005542128928937018,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0004029279872383995,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.562499950666638,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2559.6974170002795,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 79.46579200142878,
"inferenceMs": 62.12887500078068,
"imagesPerSecond": 16.09557552727994,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2559.6974170002795,
"totalMs": 2639.1632090017083,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1493696512
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00009412708459421992,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0007523161411882029,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.500000024236441,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2559.6974170002795,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 79.47666700056288,
"inferenceMs": 65.1930829990306,
"imagesPerSecond": 15.339050616993674,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 1024,
"inputHeight": 683,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2559.6974170002795,
"totalMs": 2639.1740840008424,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1493696512
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0016382932662963867,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.05339636312328366,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 9.000000004326502,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2559.6974170002795,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 71.31483299963293,
"inferenceMs": 61.51991699880455,
"imagesPerSecond": 16.254898393628068,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2559.6974170002795,
"totalMs": 2631.0122499999125,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1493696512
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.007005747407674789,
"expectedLabel": "a red car",
"topLabel": "a red car",
"margin": 0.06753655802913272,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 8.874999939047145,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2559.6974170002795,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 75.76070799768786,
"inferenceMs": 65.00474999847938,
"imagesPerSecond": 15.383491206771696,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108017152,
"processInitMs": 2559.6974170002795,
"totalMs": 2635.4581249979674,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1493696512
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00038677407428622246,
"expectedLabel": null,
"topLabel": "a person",
"margin": 0.0005178638093639165,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": null,
"logOddsMargin": 1.5000000226721921,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2559.6974170002795,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 54.6254590008175,
"inferenceMs": 50.88245799925062,
"imagesPerSecond": 19.653138612421746,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 317,
"inputHeight": 229,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"processInitMs": 2559.6974170002795,
"totalMs": 2614.322876001097,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1493696512
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.002058379352092743,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.015884946718870196,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 6.625000088039937,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2559.6974170002795,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 53.99070900239167,
"inferenceMs": 48.37533300087671,
"imagesPerSecond": 20.671692326787227,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 512,
"inputHeight": 256,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"processInitMs": 2559.6974170002795,
"totalMs": 2613.688126002671,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1493696512
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.00015164026990532875,
"expectedLabel": "a person",
"topLabel": "a person",
"margin": 0.0008635782869532704,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 1.8125000765513555,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2559.6974170002795,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 52.99712499981979,
"inferenceMs": 48.06700000335695,
"imagesPerSecond": 20.804294004829945,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 256,
"inputHeight": 512,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"processInitMs": 2559.6974170002795,
"totalMs": 2612.6945420000993,
"initializationReused": true,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "excellent",
"observedResidentBytesAtCaseEnd": 1493696512
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0005243769846856594,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.006270976529776817,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.812499953123904,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2559.6974170002795,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 71.15283399980399,
"inferenceMs": 58.885249996819766,
"imagesPerSecond": 16.982181447034822,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2630.8502510000835,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1493696512
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0005243769846856594,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.006270976529776817,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.812499953123904,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2559.6974170002795,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 73.57374999992317,
"inferenceMs": 59.46224999934202,
"imagesPerSecond": 16.817392547558587,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2633.2711670002027,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1493696512
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0005243769846856594,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.006270976529776817,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.812499953123904,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2559.6974170002795,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 78.84533400283544,
"inferenceMs": 66.12300000051619,
"imagesPerSecond": 15.123330762249044,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2638.542751003115,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1493696512
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0005243769846856594,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.006270976529776817,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.812499953123904,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2559.6974170002795,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 75.4930000002787,
"inferenceMs": 63.0837499993504,
"imagesPerSecond": 15.851942853909247,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2635.190417000558,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1493696512
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"finite": true,
"completeLabels": true,
"cpuScoreDifference": 0.0005243769846856594,
"expectedLabel": "a dog",
"topLabel": "a dog",
"margin": 0.006270976529776817,
"scope": "Three cleared image subjects, brightness derivatives, and blank numerical stability with four English labels; not broad classification, multilingual, safety, or unknown-image rejection certification",
"top1Accuracy": 1,
"logOddsMargin": 5.812499953123904,
"passed": true
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 32,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2559.6974170002795,
"placement": {
"mps:0/torch.bfloat16": 408
},
"parameterDtype": "torch.bfloat16",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 80.22875000096974,
"inferenceMs": 68.03695800044807,
"imagesPerSecond": 14.697894047429552,
"outputPlacement": "mps:0",
"modelInputShape": [
1,
3,
256,
256
],
"spatialShapes": null,
"validPatchCount": null,
"inputWidth": 700,
"inputHeight": 577,
"labelCount": 4,
"scoreFunction": "sigmoid",
"promptTemplate": "This is a photo of {label}.",
"driverAllocatedMemoryBytes": 1108000768,
"totalMs": 2639.9261670012493,
"device": "mps",
"precision": "bf16",
"cpuThreads": 4,
"artifactRevision": "94dffa8cb1179de3e03f091dbc3917e5d5a9ae84",
"checkpoint": "siglip2-base-patch32-256",
"fallback": false,
"grade": "good",
"observedResidentBytesAtCaseEnd": 1493696512
}
}
]