SigLIP 1 Desktop
Compare still images with candidate text labels using Google SigLIP v1 So400m/16 multilingual i18n checkpoint at256pixels. Returns complete independently scored ranked labels.
Desktop app
Run SigLIP 1 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 1 So400m i18n — Image & Text Matching (256) Desktop
225 configurations testedQuality criteria
{ "frozenBeforeInference": true, "minimumTop1Accuracy": 1, "minimumEnglishLogOddsMargin": 1, "minimumMultilingualLogOddsMargin": 0.5, "maximumCpuScoreDifference": 0.03, "finiteScoresInUnitInterval": true, "maximumWallInferenceMs": { "cpu": 4000, "mps": 1500 }, "goodInferenceMs": 250, "scope": "Three cleared image subjects with brightness derivatives, blank numerical stability and French/Chinese dog label probes. Not broad accuracy, all-language coverage, calibrated confidence or unknown-object rejection. CPU image/text preparation and IPC included in wall timings.", "matrix": { "cpuThreads": [ 1, 4 ], "mpsPrecisions": [ "fp32", "fp16", "bf16" ], "coldRepetitions": 3, "warmupExcluded": 1, "warmRepeats": 5 } }
| Precision | Device | Result | Platform | Inference / load | |
|---|---|---|---|---|---|
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 83 ms / 2371 ms |
BF16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 83 ms / 2389 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 83 ms / 2287 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 83 ms / 2431 ms |
FP16 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 84 ms / 2329 ms |
FP32 | mpsAccelerated | Passed | Very fast | macos · aarch64 | 100 ms / 2705 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 101 ms / 2654 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 101 ms / 2658 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 103 ms / 2633 ms |
BF16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 118 ms / 2389 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 118 ms / 2287 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 120 ms / 2287 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 127 ms / 2287 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 128 ms / 2431 ms |
BF16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 129 ms / 2389 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 134 ms / 2705 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 134 ms / 2371 ms |
BF16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 134 ms / 2389 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 135 ms / 2431 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 136 ms / 2287 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 136 ms / 2705 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 137 ms / 2371 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 137 ms / 2287 ms |
BF16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 138 ms / 2389 ms |
BF16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 138 ms / 2389 ms |
BF16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 138 ms / 2389 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 138 ms / 2329 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 138 ms / 2287 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 139 ms / 2287 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 139 ms / 2329 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 146 ms / 2431 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 146 ms / 2431 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 147 ms / 2371 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 147 ms / 2705 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 147 ms / 2633 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 148 ms / 2658 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 148 ms / 2654 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 148 ms / 2329 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 148 ms / 2658 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 148 ms / 2371 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 149 ms / 2633 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 149 ms / 2329 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 149 ms / 2371 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 150 ms / 2371 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 150 ms / 2431 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 151 ms / 2371 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 151 ms / 2654 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 152 ms / 2705 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 152 ms / 2329 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 152 ms / 2431 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 153 ms / 2371 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 153 ms / 2705 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 153 ms / 2329 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 153 ms / 2329 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 153 ms / 2705 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 153 ms / 2431 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 153 ms / 2431 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 154 ms / 2371 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 154 ms / 2431 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 154 ms / 2371 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 155 ms / 2705 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 156 ms / 2329 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 156 ms / 2431 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 156 ms / 2329 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 156 ms / 2431 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 158 ms / 2371 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 159 ms / 2431 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 160 ms / 2654 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 160 ms / 2329 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 160 ms / 2371 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 160 ms / 2431 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 161 ms / 2654 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 162 ms / 2329 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 162 ms / 2329 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 162 ms / 2371 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 162 ms / 2633 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 163 ms / 2654 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 163 ms / 2329 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 163 ms / 2658 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 164 ms / 2654 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 166 ms / 2654 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 166 ms / 2654 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 166 ms / 2654 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 166 ms / 2633 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 166 ms / 2658 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 167 ms / 2633 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 167 ms / 2658 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 168 ms / 2633 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 168 ms / 2654 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 168 ms / 2633 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 168 ms / 2654 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 169 ms / 2658 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 169 ms / 2633 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 169 ms / 2658 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 169 ms / 2633 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 169 ms / 2658 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 170 ms / 2654 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 170 ms / 2658 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 170 ms / 2654 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 170 ms / 2658 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 170 ms / 2633 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 170 ms / 2658 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 170 ms / 2633 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 170 ms / 2658 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 171 ms / 2633 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 176 ms / 2633 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 185 ms / 2658 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 274 ms / 2654 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 278 ms / 2658 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 280 ms / 2371 ms |
FP32 | mpsAccelerated | Passed | Fast | macos · aarch64 | 319 ms / 2633 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 321 ms / 2329 ms |
FP16 | mpsAccelerated | Passed | Fast | macos · aarch64 | 335 ms / 2431 ms |
FP32 | mpsAccelerated | Passed | Moderate | macos · aarch64 | 1015 ms / 2705 ms |
FP16 | mpsAccelerated | Passed | Moderate | macos · aarch64 | 1043 ms / 2287 ms |
BF16 | mpsAccelerated | Passed | Moderate | macos · aarch64 | 1063 ms / 2389 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 347 ms / 1933 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 348 ms / 1929 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 348 ms / 1932 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 350 ms / 2010 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 380 ms / 1932 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 382 ms / 1933 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 382 ms / 1933 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 382 ms / 1932 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 383 ms / 1929 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 383 ms / 1929 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 384 ms / 2010 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 392 ms / 2010 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 392 ms / 1932 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 393 ms / 1929 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 394 ms / 1933 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 396 ms / 1932 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 397 ms / 1932 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 398 ms / 1932 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 398 ms / 1929 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 398 ms / 1933 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 399 ms / 1929 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 400 ms / 2010 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 400 ms / 2010 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 400 ms / 2010 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 400 ms / 1933 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 400 ms / 1929 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 400 ms / 2010 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 401 ms / 2010 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 401 ms / 2010 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 401 ms / 1932 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 401 ms / 1933 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 401 ms / 1932 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 401 ms / 1929 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 402 ms / 2010 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 402 ms / 1929 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 402 ms / 1929 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 402 ms / 1929 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 402 ms / 1932 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 403 ms / 2010 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 404 ms / 2010 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 404 ms / 1932 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 407 ms / 1932 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 408 ms / 2010 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 408 ms / 1929 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 409 ms / 1932 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 413 ms / 1929 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 430 ms / 1932 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 430 ms / 2010 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 431 ms / 1929 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 431 ms / 1933 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 550 ms / 2891 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 555 ms / 1934 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 556 ms / 1964 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 567 ms / 1912 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 583 ms / 2891 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 588 ms / 2891 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 589 ms / 1964 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 589 ms / 1934 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 590 ms / 1912 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 590 ms / 1934 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 599 ms / 1912 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 600 ms / 1964 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 601 ms / 1934 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 602 ms / 1912 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 602 ms / 1964 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 603 ms / 2891 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 604 ms / 1934 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 604 ms / 1934 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 604 ms / 1964 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 604 ms / 1964 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 605 ms / 1934 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 605 ms / 1964 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 605 ms / 1912 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 605 ms / 1964 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 606 ms / 1912 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 606 ms / 2891 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 606 ms / 2891 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 606 ms / 1912 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 606 ms / 1912 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 606 ms / 1964 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 606 ms / 1912 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 607 ms / 1934 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 607 ms / 1964 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 607 ms / 1912 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 607 ms / 1934 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 607 ms / 1912 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 608 ms / 1934 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 608 ms / 1912 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 609 ms / 1964 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 609 ms / 1912 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 610 ms / 1934 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 610 ms / 1964 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 612 ms / 2891 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 617 ms / 1964 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 621 ms / 1934 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 623 ms / 1934 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 636 ms / 1964 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 638 ms / 1912 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 640 ms / 1934 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 643 ms / 2891 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 706 ms / 1929 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 709 ms / 1933 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 712 ms / 2010 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 713 ms / 1932 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 912 ms / 1912 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 918 ms / 1964 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 922 ms / 1934 ms |
FP32 | cpu | Passed | Fast | macos · aarch64 | 940 ms / 2891 ms |
BF16 | mps | Failed | — | macos · aarch64 | 121 ms / 2389 ms |
Test setup & sample details
- Quality gate
- { "frozenBeforeInference": true, "minimumTop1Accuracy": 1, "minimumEnglishLogOddsMargin": 1, "minimumMultilingualLogOddsMargin": 0.5, "maximumCpuScoreDifference": 0.03, "finiteScoresInUnitInterval": true, "maximumWallInferenceMs": { "cpu": 4000, "mps": 1500 }, "goodInferenceMs": 250, "scope": "Three cleared image subjects with brightness derivatives, blank numerical stability and French/Chinese dog label probes. Not broad accuracy, all-language coverage, calibrated confidence or unknown-object rejection. CPU image/text preparation and IPC included in wall timings.", "matrix": { "cpuThreads": [ 1, 4 ], "mpsPrecisions": [ "fp32", "fp16", "bf16" ], "coldRepetitions": 3, "warmupExcluded": 1, "warmRepeats": 5 } }
- Tested
- 2026-11-13T12:06:32.570046+00:00
- Environment
- { "hardware": "Apple M3Max;Mac15,11;96GiB", "os": "macOS27.2", "runtime": "Transformers5.22.1/Torch2.10.0/Pillow12.0.0/SentencePiece0.2.1/Protobuf6.33.5", "interpreterScope": "Existing isolated app-owned development interpreter plus model-owned pinned tokenizer overlay; normal managed provisioning is separate functional evidence.", "transport": "Hardlinked checksum-verified original local snapshot, not Internet transfer timing", "concurrentUploadActivity": "none; excluded", "concurrentTaskBuildInferenceDownload": "none during speed cases" }
- Runtime
- transformers 5.22.1
- Artifact revision
- 365d321c0cfdea96bc28e3a29787a11a062681a1
- 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": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2891.2095419946127,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 893.2234160020016,
"inferenceMs": 940.0604160036892,
"imagesPerSecond": 1.063761416793956,
"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,
"modelInferenceMs": 864.6564160007983,
"totalMs": 3831.269957998302,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3803119616,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 10.95101885160621,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2891.2095419946127,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 595.111666014418,
"inferenceMs": 642.8207080462016,
"imagesPerSecond": 1.5556437237988399,
"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,
"modelInferenceMs": 583.617708995007,
"totalMs": 3534.0302500408143,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3808772096,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.524991987271818,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2891.2095419946127,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 565.3261250117794,
"inferenceMs": 603.0188329750672,
"imagesPerSecond": 1.6583229997417785,
"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,
"modelInferenceMs": 551.9526660209522,
"totalMs": 3494.22837496968,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3827220480,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.807835953516252,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2891.2095419946127,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 569.5459580165334,
"inferenceMs": 611.6727910120972,
"imagesPerSecond": 1.6348610150622556,
"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,
"modelInferenceMs": 555.499707988929,
"totalMs": 3502.88233300671,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3829530624,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.698087760556987,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2891.2095419946127,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 555.7697080075741,
"inferenceMs": 588.3891670382582,
"imagesPerSecond": 1.6995554235535035,
"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,
"modelInferenceMs": 547.1680840128101,
"totalMs": 3479.598709032871,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3829235712,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.76818182179674,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2891.2095419946127,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 551.907500019297,
"inferenceMs": 583.1806249916553,
"imagesPerSecond": 1.714734607334235,
"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,
"modelInferenceMs": 543.5315830400214,
"totalMs": 3474.390166986268,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3832315904,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": null,
"top1": "a person",
"logOddsMargin": null,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2891.2095419946127,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 548.2535000192001,
"inferenceMs": 549.7761669685133,
"imagesPerSecond": 1.8189220633445027,
"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,
"modelInferenceMs": 545.9216250455938,
"totalMs": 3440.985708963126,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3832332288,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "un chien",
"top1": "un chien",
"logOddsMargin": 9.700699366827896,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2891.2095419946127,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 560.4496250161901,
"inferenceMs": 605.8056250330992,
"imagesPerSecond": 1.650694478027277,
"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,
"modelInferenceMs": 549.4619590463117,
"totalMs": 3497.015167027712,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3834478592,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "一只狗",
"top1": "一只狗",
"logOddsMargin": 2.790970286157167,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2891.2095419946127,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 560.6217919848859,
"inferenceMs": 605.9013749472797,
"imagesPerSecond": 1.6504336206317594,
"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,
"modelInferenceMs": 549.4590420275927,
"totalMs": 3497.1109169418924,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3838820352,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1933.1573329982348,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 663.0152500001714,
"inferenceMs": 708.9011250063777,
"imagesPerSecond": 1.4106339582843848,
"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,
"modelInferenceMs": 645.3572089667432,
"totalMs": 2642.0584580046125,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3802857472,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 10.95101885160621,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1933.1573329982348,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 385.52791700931266,
"inferenceMs": 431.1238750233315,
"imagesPerSecond": 2.3195189548384025,
"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,
"modelInferenceMs": 374.41512499935925,
"totalMs": 2364.2812080215663,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3808608256,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.524991987271818,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1933.1573329982348,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 357.78224997920915,
"inferenceMs": 393.57820799341425,
"imagesPerSecond": 2.5407910796136686,
"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,
"modelInferenceMs": 344.55700003309175,
"totalMs": 2326.735540991649,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3818651648,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.807835953516252,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1933.1573329982348,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 357.0524589740671,
"inferenceMs": 398.364957945887,
"imagesPerSecond": 2.5102609555728987,
"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,
"modelInferenceMs": 343.76833302667364,
"totalMs": 2331.522290944122,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3820961792,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.698087760556987,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1933.1573329982348,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 351.74983396427706,
"inferenceMs": 382.15191703056917,
"imagesPerSecond": 2.6167603914440334,
"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,
"modelInferenceMs": 343.7592500122264,
"totalMs": 2315.309250028804,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3820699648,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.76818182179674,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1933.1573329982348,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 350.713416992221,
"inferenceMs": 381.63233298109844,
"imagesPerSecond": 2.620323053313012,
"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,
"modelInferenceMs": 342.733874975238,
"totalMs": 2314.789665979333,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3823779840,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": null,
"top1": "a person",
"logOddsMargin": null,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1933.1573329982348,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 345.1525840209797,
"inferenceMs": 346.6646669548936,
"imagesPerSecond": 2.884632024324865,
"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,
"modelInferenceMs": 342.84641599515453,
"totalMs": 2279.8219999531284,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3823828992,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "un chien",
"top1": "un chien",
"logOddsMargin": 9.700699366827896,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1933.1573329982348,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 354.48266699677333,
"inferenceMs": 400.24483401793987,
"imagesPerSecond": 2.4984707234351906,
"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,
"modelInferenceMs": 343.4163749916479,
"totalMs": 2333.4021670161746,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3825991680,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "一只狗",
"top1": "一只狗",
"logOddsMargin": 2.790970286157167,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1933.1573329982348,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 355.1380829885602,
"inferenceMs": 401.4003339689225,
"imagesPerSecond": 2.491278445417095,
"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,
"modelInferenceMs": 344.06079194741324,
"totalMs": 2334.5576669671573,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3830317056,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2705.494499998167,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 968.7045839964412,
"inferenceMs": 1015.1042499928735,
"imagesPerSecond": 0.9851204937887122,
"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": 5589745664,
"modelInferenceMs": 933.8553340057842,
"totalMs": 3720.5987499910407,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 497795072,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "acceptable",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 10.951022176258594,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 5.4016709327697754e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2705.494499998167,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 109.41629199078307,
"inferenceMs": 155.0817089737393,
"imagesPerSecond": 6.448213697266739,
"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": 5589745664,
"modelInferenceMs": 96.07704199152067,
"totalMs": 2860.5762089719065,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 501055488,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.524990572425493,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 1.30385160446167e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2705.494499998167,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 111.00812500808388,
"inferenceMs": 146.70933299930766,
"imagesPerSecond": 6.816198939468419,
"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": 5589745664,
"modelInferenceMs": 96.21225000591949,
"totalMs": 2852.203832997475,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 514015232,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.807834057743548,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 1.1827796697616577e-7,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2705.494499998167,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 110.48949998803437,
"inferenceMs": 151.83279098710045,
"imagesPerSecond": 6.5861925707797795,
"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": 5589745664,
"modelInferenceMs": 95.63979203812778,
"totalMs": 2857.3272909852676,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 519421952,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.698091432457698,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000004231929779052734,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2705.494499998167,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 104.49183400487527,
"inferenceMs": 136.4161660312675,
"imagesPerSecond": 7.3305094923338725,
"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": 5589745664,
"modelInferenceMs": 95.27808398706838,
"totalMs": 2841.9106660294347,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 520519680,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.76815894415141,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 8.940696716308594e-7,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2705.494499998167,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 104.3089579907246,
"inferenceMs": 133.56216700049117,
"imagesPerSecond": 7.487150159792799,
"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": 5589745664,
"modelInferenceMs": 94.67900003073737,
"totalMs": 2839.0566669986583,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 521453568,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": null,
"top1": "a person",
"logOddsMargin": null,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 7.639755494892597e-10,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2705.494499998167,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 98.3172919950448,
"inferenceMs": 99.64008297538385,
"imagesPerSecond": 10.03612171064782,
"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": 5589745664,
"modelInferenceMs": 95.02854198217392,
"totalMs": 2805.134582973551,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 521469952,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "un chien",
"top1": "un chien",
"logOddsMargin": 9.700717012504292,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 7.82310962677002e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2705.494499998167,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 106.85762501088902,
"inferenceMs": 152.6279579848051,
"imagesPerSecond": 6.5518795717594225,
"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": 5589745664,
"modelInferenceMs": 94.0958330174908,
"totalMs": 2858.1224579829723,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 524271616,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "一只狗",
"top1": "一只狗",
"logOddsMargin": 2.7909704311167927,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 8.381903171539307e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2705.494499998167,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 108.07408299297094,
"inferenceMs": 153.28416600823402,
"imagesPerSecond": 6.5238310390538485,
"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": 5589745664,
"modelInferenceMs": 95.58637498412281,
"totalMs": 2858.778666006401,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 526237696,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2287.00812504394,
"placement": {
"mps:0/torch.float16": 888
},
"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": 996.8154160305858,
"inferenceMs": 1043.0794170242734,
"imagesPerSecond": 0.9586997726911614,
"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": 2372714496,
"modelInferenceMs": 967.4468750017695,
"totalMs": 3330.087542068213,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 833306624,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "acceptable",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 10.957031159204348,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.00016664713621139526,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2287.00812504394,
"placement": {
"mps:0/torch.float16": 888
},
"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": 91.15916694281623,
"inferenceMs": 138.2164999959059,
"imagesPerSecond": 7.235026209096749,
"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": 2372714496,
"modelInferenceMs": 78.37433298118412,
"totalMs": 2425.2246250398457,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 844742656,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.546875029154597,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.00011055544018745422,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2287.00812504394,
"placement": {
"mps:0/torch.float16": 888
},
"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": 91.7910419520922,
"inferenceMs": 126.81829096982256,
"imagesPerSecond": 7.885297872670104,
"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": 2372714496,
"modelInferenceMs": 77.63570797396824,
"totalMs": 2413.8264160137624,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 846856192,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.828124881494531,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000033278949558734894,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2287.00812504394,
"placement": {
"mps:0/torch.float16": 888
},
"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": 93.3078330126591,
"inferenceMs": 136.23462500981987,
"imagesPerSecond": 7.340277847338145,
"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": 2372714496,
"modelInferenceMs": 78.42462498228997,
"totalMs": 2423.2427500537597,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 849330176,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.710937433877344,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.0005918145179748535,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2287.00812504394,
"placement": {
"mps:0/torch.float16": 888
},
"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": 88.00770802190527,
"inferenceMs": 118.07491700164974,
"imagesPerSecond": 8.46919926258367,
"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": 2372714496,
"modelInferenceMs": 77.63987500220537,
"totalMs": 2405.0830420455895,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 850804736,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.781250162637669,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.00009149312973022461,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2287.00812504394,
"placement": {
"mps:0/torch.float16": 888
},
"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": 88.30937498714775,
"inferenceMs": 119.94500004220754,
"imagesPerSecond": 8.33715452622543,
"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": 2372714496,
"modelInferenceMs": 78.74154200544581,
"totalMs": 2406.9531250861473,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 851902464,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": null,
"top1": "a person",
"logOddsMargin": null,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 4.491812433116138e-7,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2287.00812504394,
"placement": {
"mps:0/torch.float16": 888
},
"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": 81.98000001721084,
"inferenceMs": 83.40733300428838,
"imagesPerSecond": 11.989353501431165,
"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": 2372714496,
"modelInferenceMs": 78.58712499728426,
"totalMs": 2370.415458048228,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 846626816,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "un chien",
"top1": "un chien",
"logOddsMargin": 9.69531245881724,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0.00006689317524433136,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2287.00812504394,
"placement": {
"mps:0/torch.float16": 888
},
"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": 91.47483296692371,
"inferenceMs": 138.8854999677278,
"imagesPerSecond": 7.200175685959769,
"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": 2372714496,
"modelInferenceMs": 78.55462504085153,
"totalMs": 2425.8936250116676,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 850919424,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "一只狗",
"top1": "一只狗",
"logOddsMargin": 2.7851561236948346,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0.00002637971192598343,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2287.00812504394,
"placement": {
"mps:0/torch.float16": 888
},
"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": 90.36491601727903,
"inferenceMs": 137.40541599690914,
"imagesPerSecond": 7.27773350667993,
"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": 2372714496,
"modelInferenceMs": 78.0287500238046,
"totalMs": 2424.413541040849,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 853983232,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.187499943558505,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.001644376665353775,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2388.6759169981815,
"placement": {
"mps:0/torch.bfloat16": 888
},
"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": 1019.136291986797,
"inferenceMs": 1063.0213750409894,
"imagesPerSecond": 0.9407148562383713,
"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": 2372714496,
"modelInferenceMs": 988.7284589931369,
"totalMs": 3451.697292039171,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "bf16",
"actualPrecision": "bf16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 832077824,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "acceptable",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 10.99999994237012,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.00007879175245761871,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2388.6759169981815,
"placement": {
"mps:0/torch.bfloat16": 888
},
"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": 92.01029204996303,
"inferenceMs": 137.7903749817051,
"imagesPerSecond": 7.257400962387782,
"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": 2372714496,
"modelInferenceMs": 79.06616700347513,
"totalMs": 2526.4662919798866,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "bf16",
"actualPrecision": "bf16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 842334208,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.624999991266979,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.00040644221007823944,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2388.6759169981815,
"placement": {
"mps:0/torch.bfloat16": 888
},
"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": 93.61287497449666,
"inferenceMs": 128.51408403366804,
"imagesPerSecond": 7.78124831623918,
"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": 2372714496,
"modelInferenceMs": 78.9282500045374,
"totalMs": 2517.1900010318495,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "bf16",
"actualPrecision": "bf16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 844595200,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.843749920896705,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.0006674434989690781,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2388.6759169981815,
"placement": {
"mps:0/torch.bfloat16": 888
},
"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": 93.22687500389293,
"inferenceMs": 134.2396250111051,
"imagesPerSecond": 7.449365266904419,
"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": 2372714496,
"modelInferenceMs": 78.12662498326972,
"totalMs": 2522.9155420092866,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "bf16",
"actualPrecision": "bf16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 847069184,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "bf16",
"device": "mps",
"status": "fail",
"quality": {
"passed": false,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.812500176004358,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.039862215518951416,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2388.6759169981815,
"placement": {
"mps:0/torch.bfloat16": 888
},
"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": 88.90187501674518,
"inferenceMs": 120.63541600946337,
"imagesPerSecond": 8.289439644503352,
"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": 2372714496,
"modelInferenceMs": 79.08845803467557,
"totalMs": 2509.311333007645,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "bf16",
"actualPrecision": "bf16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 849936384,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.81250020489218,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.0017694830894470215,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2388.6759169981815,
"placement": {
"mps:0/torch.bfloat16": 888
},
"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": 87.27758302120492,
"inferenceMs": 117.7802499732934,
"imagesPerSecond": 8.490387821614824,
"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": 2372714496,
"modelInferenceMs": 78.23820802150294,
"totalMs": 2506.456166971475,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "bf16",
"actualPrecision": "bf16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 851034112,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": null,
"top1": "a person",
"logOddsMargin": null,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000004450826963875443,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2388.6759169981815,
"placement": {
"mps:0/torch.bfloat16": 888
},
"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": 81.78929198766127,
"inferenceMs": 83.22200004477054,
"imagesPerSecond": 12.016053440941516,
"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": 2372714496,
"modelInferenceMs": 78.31258396618068,
"totalMs": 2471.897917042952,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "bf16",
"actualPrecision": "bf16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 845774848,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "un chien",
"top1": "un chien",
"logOddsMargin": 9.781250007486587,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0.0001144707202911377,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2388.6759169981815,
"placement": {
"mps:0/torch.bfloat16": 888
},
"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": 92.0656660455279,
"inferenceMs": 137.77558400761336,
"imagesPerSecond": 7.258180084685693,
"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": 2372714496,
"modelInferenceMs": 78.99537496268749,
"totalMs": 2526.451501005795,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "bf16",
"actualPrecision": "bf16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 850100224,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "bf16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "一只狗",
"top1": "一只狗",
"logOddsMargin": 2.8125000974370487,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0.00006081443279981613,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2388.6759169981815,
"placement": {
"mps:0/torch.bfloat16": 888
},
"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": 91.7520000366494,
"inferenceMs": 137.61074998183176,
"imagesPerSecond": 7.266874136882666,
"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": 2372714496,
"modelInferenceMs": 78.71983299264684,
"totalMs": 2526.2866669800133,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "bf16",
"actualPrecision": "bf16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 853164032,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1934.237374982331,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 877.1770000457764,
"inferenceMs": 921.6557500185445,
"imagesPerSecond": 1.0850038096977956,
"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,
"modelInferenceMs": 859.5142909907736,
"totalMs": 2855.8931250008754,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3804610560,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1934.237374982331,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 594.4358329870738,
"inferenceMs": 640.3692500316538,
"imagesPerSecond": 1.5615990304821312,
"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,
"modelInferenceMs": 583.3659580093808,
"totalMs": 2574.6066250139847,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3810721792,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 10.95101885160621,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1934.237374982331,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 561.0888340161182,
"inferenceMs": 604.1999589651823,
"imagesPerSecond": 1.6550812113802644,
"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,
"modelInferenceMs": 550.0122500234284,
"totalMs": 2538.4373339475133,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3818749952,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.524991987271818,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1934.237374982331,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 563.6123339645565,
"inferenceMs": 600.8213750319555,
"imagesPerSecond": 1.6643881884974774,
"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,
"modelInferenceMs": 550.431750016287,
"totalMs": 2535.0587500142865,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3828776960,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.807835953516252,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1934.237374982331,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 562.9731660010293,
"inferenceMs": 604.0317909792066,
"imagesPerSecond": 1.6555420011567312,
"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,
"modelInferenceMs": 549.638208001852,
"totalMs": 2538.2691659615375,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3833217024,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.698087760556987,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1934.237374982331,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 557.8212499967776,
"inferenceMs": 589.4106249907054,
"imagesPerSecond": 1.6966100670746636,
"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,
"modelInferenceMs": 549.7716669924557,
"totalMs": 2523.6479999730363,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3832922112,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.76818182179674,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1934.237374982331,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 559.7572079859674,
"inferenceMs": 590.2797500020824,
"imagesPerSecond": 1.6941119867257384,
"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,
"modelInferenceMs": 551.7240419867449,
"totalMs": 2524.5171249844134,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3836002304,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": null,
"top1": "a person",
"logOddsMargin": null,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1934.237374982331,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 553.3649999997579,
"inferenceMs": 554.8777919611894,
"imagesPerSecond": 1.8021986363259326,
"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,
"modelInferenceMs": 551.1115419794805,
"totalMs": 2489.1151669435203,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3835985920,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "un chien",
"top1": "un chien",
"logOddsMargin": 9.700699366827896,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1934.237374982331,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 576.8134579993784,
"inferenceMs": 622.8097499697469,
"imagesPerSecond": 1.6056267584901731,
"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,
"modelInferenceMs": 565.2438329998404,
"totalMs": 2557.047124952078,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3838181376,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "一只狗",
"top1": "一只狗",
"logOddsMargin": 2.790970286157167,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1934.237374982331,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 573.9899579784833,
"inferenceMs": 620.8848750102334,
"imagesPerSecond": 1.6106045424017104,
"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,
"modelInferenceMs": 562.2981670312583,
"totalMs": 2555.1222499925643,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3842506752,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1934.237374982331,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 561.5679589682259,
"inferenceMs": 607.5450840289705,
"imagesPerSecond": 1.645968383726261,
"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,
"modelInferenceMs": 550.4104999708943,
"totalMs": 2541.7824590113014,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3848208384,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1934.237374982331,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 561.7310000234284,
"inferenceMs": 607.1668750373647,
"imagesPerSecond": 1.6469936702960954,
"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,
"modelInferenceMs": 550.5501250154339,
"totalMs": 2541.4042500196956,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3853910016,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1934.237374982331,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 561.0557909822091,
"inferenceMs": 606.6747500444762,
"imagesPerSecond": 1.6483296855962581,
"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,
"modelInferenceMs": 549.9344579875469,
"totalMs": 2540.912125026807,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3859611648,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1934.237374982331,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 560.5879590148106,
"inferenceMs": 604.6849580015987,
"imagesPerSecond": 1.6537537221115333,
"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,
"modelInferenceMs": 549.4108750135638,
"totalMs": 2538.9223329839297,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3865313280,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1934.237374982331,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 561.9132090359926,
"inferenceMs": 609.7022499889135,
"imagesPerSecond": 1.6401448412207489,
"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,
"modelInferenceMs": 550.7184590096585,
"totalMs": 2543.9396249712445,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3874570240,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1912.3099169810303,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 865.0618749670684,
"inferenceMs": 912.1072919806466,
"imagesPerSecond": 1.0963622468454275,
"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,
"modelInferenceMs": 847.211625019554,
"totalMs": 2824.417208961677,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3804905472,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1912.3099169810303,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 592.2987499507144,
"inferenceMs": 638.194749946706,
"imagesPerSecond": 1.5669198157513948,
"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,
"modelInferenceMs": 581.2482089968398,
"totalMs": 2550.5046669277363,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3811033088,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 10.95101885160621,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1912.3099169810303,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 562.0661660213955,
"inferenceMs": 606.9247919949703,
"imagesPerSecond": 1.6476506038136718,
"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,
"modelInferenceMs": 551.0129579924978,
"totalMs": 2519.2347089760005,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3816685568,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.524991987271818,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1912.3099169810303,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 563.3404580294155,
"inferenceMs": 599.1175000090152,
"imagesPerSecond": 1.6691216664259558,
"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,
"modelInferenceMs": 550.1174579840153,
"totalMs": 2511.4274169900455,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3826745344,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.807835953516252,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1912.3099169810303,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 566.3997079827823,
"inferenceMs": 608.1437499960884,
"imagesPerSecond": 1.6443480673877384,
"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,
"modelInferenceMs": 552.8189169708639,
"totalMs": 2520.4536669771187,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3831185408,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.698087760556987,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1912.3099169810303,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 557.7883329824544,
"inferenceMs": 589.7424999857321,
"imagesPerSecond": 1.6956553072301783,
"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,
"modelInferenceMs": 549.8112910427153,
"totalMs": 2502.0524169667624,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3830890496,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.76818182179674,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1912.3099169810303,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 572.3845830070786,
"inferenceMs": 601.7528750235215,
"imagesPerSecond": 1.6618117528078478,
"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,
"modelInferenceMs": 564.0882499865256,
"totalMs": 2514.0627920045517,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3833970688,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": null,
"top1": "a person",
"logOddsMargin": null,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1912.3099169810303,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 565.6774580129422,
"inferenceMs": 567.267082980834,
"imagesPerSecond": 1.7628380528361922,
"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,
"modelInferenceMs": 563.2298749987967,
"totalMs": 2479.576999961864,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3833987072,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "un chien",
"top1": "un chien",
"logOddsMargin": 9.700699366827896,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1912.3099169810303,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 560.8046249835752,
"inferenceMs": 608.8435410056263,
"imagesPerSecond": 1.6424580908722477,
"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,
"modelInferenceMs": 549.6074589900672,
"totalMs": 2521.1534579866566,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3842457600,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "一只狗",
"top1": "一只狗",
"logOddsMargin": 2.790970286157167,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1912.3099169810303,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 561.262707982678,
"inferenceMs": 605.2092080353759,
"imagesPerSecond": 1.6523211919497887,
"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,
"modelInferenceMs": 550.0502500217408,
"totalMs": 2517.519125016406,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3846782976,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1912.3099169810303,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 559.6312080160715,
"inferenceMs": 605.5354590062052,
"imagesPerSecond": 1.6514309527656457,
"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,
"modelInferenceMs": 548.3468329766765,
"totalMs": 2517.8453759872355,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3852484608,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1912.3099169810303,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 560.9374580089934,
"inferenceMs": 607.3629170423374,
"imagesPerSecond": 1.646462060722573,
"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,
"modelInferenceMs": 549.8272080440074,
"totalMs": 2519.6728340233676,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3858186240,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1912.3099169810303,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 560.459083004389,
"inferenceMs": 606.2505000154488,
"imagesPerSecond": 1.649483175641946,
"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,
"modelInferenceMs": 549.2914170026779,
"totalMs": 2518.560416996479,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3863887872,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1912.3099169810303,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 559.6963749849238,
"inferenceMs": 605.9438329539262,
"imagesPerSecond": 1.6503179760491704,
"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,
"modelInferenceMs": 548.4302919940092,
"totalMs": 2518.2537499349564,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3869589504,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1912.3099169810303,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 560.381333983969,
"inferenceMs": 605.9195000561886,
"imagesPerSecond": 1.6503842505601276,
"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,
"modelInferenceMs": 549.1267910110764,
"totalMs": 2518.229417037219,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3875291136,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1963.6064999504015,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 872.3180000088178,
"inferenceMs": 918.0642500286922,
"imagesPerSecond": 1.089248383180967,
"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,
"modelInferenceMs": 854.7500839922577,
"totalMs": 2881.6707499790937,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3804348416,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1963.6064999504015,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 592.4325839732774,
"inferenceMs": 636.3056249683723,
"imagesPerSecond": 1.5715718371179026,
"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,
"modelInferenceMs": 581.4284589723684,
"totalMs": 2599.9121249187738,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3812802560,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 10.95101885160621,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1963.6064999504015,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 561.1552500049584,
"inferenceMs": 604.232499958016,
"imagesPerSecond": 1.654992076840427,
"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,
"modelInferenceMs": 550.1122500281781,
"totalMs": 2567.8389999084175,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3818455040,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.524991987271818,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1963.6064999504015,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 562.4519999837503,
"inferenceMs": 599.5516669936478,
"imagesPerSecond": 1.6679129673916742,
"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,
"modelInferenceMs": 549.1794580011629,
"totalMs": 2563.1581669440493,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3828498432,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.807835953516252,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1963.6064999504015,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 576.0407500201836,
"inferenceMs": 617.2547499882057,
"imagesPerSecond": 1.6200766377563034,
"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,
"modelInferenceMs": 562.0757500291802,
"totalMs": 2580.861249938607,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3833004032,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.698087760556987,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1963.6064999504015,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 570.94116701046,
"inferenceMs": 602.3040419677272,
"imagesPerSecond": 1.660291032968997,
"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,
"modelInferenceMs": 562.451166973915,
"totalMs": 2565.9105419181287,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3832741888,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.76818182179674,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1963.6064999504015,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 558.12566698296,
"inferenceMs": 589.1158750164323,
"imagesPerSecond": 1.697458925159854,
"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,
"modelInferenceMs": 549.9943339964375,
"totalMs": 2552.7223749668337,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3835822080,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": null,
"top1": "a person",
"logOddsMargin": null,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1963.6064999504015,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 554.2284170514904,
"inferenceMs": 555.7592079858296,
"imagesPerSecond": 1.799340407915468,
"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,
"modelInferenceMs": 551.9766249926761,
"totalMs": 2519.365707936231,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3835854848,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "un chien",
"top1": "un chien",
"logOddsMargin": 9.700699366827896,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1963.6064999504015,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 560.6379169621505,
"inferenceMs": 604.3978749657981,
"imagesPerSecond": 1.6545392388359876,
"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,
"modelInferenceMs": 549.6017500408925,
"totalMs": 2568.0043749161996,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3838033920,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "一只狗",
"top1": "一只狗",
"logOddsMargin": 2.790970286157167,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1963.6064999504015,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 561.0634590266272,
"inferenceMs": 606.8361250218004,
"imagesPerSecond": 1.6478913478726656,
"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,
"modelInferenceMs": 549.9691250151955,
"totalMs": 2570.442624972202,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3842375680,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1963.6064999504015,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 561.9022499886341,
"inferenceMs": 606.0854999814183,
"imagesPerSecond": 1.649932229084277,
"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,
"modelInferenceMs": 550.8614999707788,
"totalMs": 2569.6919999318197,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3848077312,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1963.6064999504015,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 561.5302079822868,
"inferenceMs": 605.092833051458,
"imagesPerSecond": 1.6526389760014866,
"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,
"modelInferenceMs": 549.8806670075282,
"totalMs": 2568.6993330018595,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3853778944,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1963.6064999504015,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 564.4091250142083,
"inferenceMs": 610.3034999687225,
"imagesPerSecond": 1.6385290270353179,
"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,
"modelInferenceMs": 553.32729103975,
"totalMs": 2573.909999919124,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3862249472,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1963.6064999504015,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 564.3760410021059,
"inferenceMs": 608.791749982629,
"imagesPerSecond": 1.6425978177735385,
"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,
"modelInferenceMs": 552.9323749942705,
"totalMs": 2572.3982499330305,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3867951104,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1963.6064999504015,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 561.1098329536617,
"inferenceMs": 605.2190000191331,
"imagesPerSecond": 1.652294458647839,
"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,
"modelInferenceMs": 550.0214170315303,
"totalMs": 2568.8254999695346,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 1,
"residentMemoryBytes": 3874832384,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2010.2224170113914,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 668.5967500088736,
"inferenceMs": 712.1657909592614,
"imagesPerSecond": 1.4041674181696322,
"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,
"modelInferenceMs": 649.9344579642639,
"totalMs": 2722.3882079706527,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3802546176,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2010.2224170113914,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 387.9480410250835,
"inferenceMs": 430.4957499844022,
"imagesPerSecond": 2.3229033040076055,
"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,
"modelInferenceMs": 376.88991596223786,
"totalMs": 2440.7181669957936,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3808722944,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 10.95101885160621,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2010.2224170113914,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 354.0492500178516,
"inferenceMs": 399.5227090199478,
"imagesPerSecond": 2.502986632357038,
"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,
"modelInferenceMs": 343.05145801045,
"totalMs": 2409.745126031339,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3814391808,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.524991987271818,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2010.2224170113914,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 364.86541596241295,
"inferenceMs": 404.0465419529937,
"imagesPerSecond": 2.4749624020203562,
"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,
"modelInferenceMs": 350.92087503289804,
"totalMs": 2414.268958964385,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3824418816,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.807835953516252,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2010.2224170113914,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 366.54733400791883,
"inferenceMs": 407.5249169836752,
"imagesPerSecond": 2.4538376877702874,
"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,
"modelInferenceMs": 352.65279200393707,
"totalMs": 2417.7473339950666,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3828973568,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.698087760556987,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2010.2224170113914,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 359.1813750099391,
"inferenceMs": 391.59466698765755,
"imagesPerSecond": 2.55366092621358,
"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,
"modelInferenceMs": 350.8484999765642,
"totalMs": 2401.817083999049,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3828711424,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.76818182179674,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2010.2224170113914,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 354.2203329852782,
"inferenceMs": 383.9304579887539,
"imagesPerSecond": 2.60463836403751,
"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,
"modelInferenceMs": 345.80749995075166,
"totalMs": 2394.1528750001453,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3831824384,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": null,
"top1": "a person",
"logOddsMargin": null,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2010.2224170113914,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 348.6808339948766,
"inferenceMs": 350.31558299669996,
"imagesPerSecond": 2.8545689901822615,
"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,
"modelInferenceMs": 346.36287501780316,
"totalMs": 2360.5380000080913,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3833085952,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "un chien",
"top1": "un chien",
"logOddsMargin": 9.700699366827896,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2010.2224170113914,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 356.7320419824682,
"inferenceMs": 403.45454099588096,
"imagesPerSecond": 2.4785939886352883,
"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,
"modelInferenceMs": 345.49883397994563,
"totalMs": 2413.6769580072723,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3835265024,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "一只狗",
"top1": "一只狗",
"logOddsMargin": 2.790970286157167,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2010.2224170113914,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 355.0596669665538,
"inferenceMs": 400.825375050772,
"imagesPerSecond": 2.494852028450872,
"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,
"modelInferenceMs": 343.86195795377716,
"totalMs": 2411.0477920621634,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3839590400,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2010.2224170113914,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 355.1513749989681,
"inferenceMs": 400.6372500443831,
"imagesPerSecond": 2.4960235222491636,
"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,
"modelInferenceMs": 344.1400419687852,
"totalMs": 2410.8596670557745,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3845259264,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2010.2224170113914,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 355.04533298080787,
"inferenceMs": 400.4045000183396,
"imagesPerSecond": 2.497474428869299,
"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,
"modelInferenceMs": 343.9098329981789,
"totalMs": 2410.626917029731,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3850960896,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2010.2224170113914,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 354.2409999645315,
"inferenceMs": 399.6476670145057,
"imagesPerSecond": 2.502204022533938,
"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,
"modelInferenceMs": 343.1355830398388,
"totalMs": 2409.870084025897,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3856662528,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2010.2224170113914,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 354.67495798366144,
"inferenceMs": 399.924999976065,
"imagesPerSecond": 2.5004688380567575,
"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,
"modelInferenceMs": 343.595375015866,
"totalMs": 2410.1474169874564,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3862364160,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2010.2224170113914,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 355.8920419891365,
"inferenceMs": 401.51208301540464,
"imagesPerSecond": 2.49058507153727,
"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,
"modelInferenceMs": 344.7831249795854,
"totalMs": 2411.734500026796,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3868065792,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1932.2110409848392,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 670.4157909844071,
"inferenceMs": 713.1536660017446,
"imagesPerSecond": 1.4022223367460804,
"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,
"modelInferenceMs": 652.8566249762662,
"totalMs": 2645.3647069865838,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3800678400,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1932.2110409848392,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 387.21562502905726,
"inferenceMs": 429.62800001259893,
"imagesPerSecond": 2.3275950356370507,
"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,
"modelInferenceMs": 376.13612500717863,
"totalMs": 2361.839040997438,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3806871552,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 10.95101885160621,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1932.2110409848392,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 354.7927920008078,
"inferenceMs": 396.8844169867225,
"imagesPerSecond": 2.5196252541037767,
"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,
"modelInferenceMs": 343.7609999673441,
"totalMs": 2329.0954579715617,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3812491264,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.524991987271818,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1932.2110409848392,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 357.20641701482236,
"inferenceMs": 392.1403330168687,
"imagesPerSecond": 2.550107489088563,
"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,
"modelInferenceMs": 344.00466695660725,
"totalMs": 2324.351374001708,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3822534656,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.807835953516252,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1932.2110409848392,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 357.7110000187531,
"inferenceMs": 396.4277089689858,
"imagesPerSecond": 2.5225280104681937,
"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,
"modelInferenceMs": 344.3113330285996,
"totalMs": 2328.638749953825,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3827892224,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.698087760556987,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1932.2110409848392,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 352.383708988782,
"inferenceMs": 382.2299580206163,
"imagesPerSecond": 2.616226119947571,
"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,
"modelInferenceMs": 344.36904202448204,
"totalMs": 2314.4409990054555,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3827875840,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.76818182179674,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1932.2110409848392,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 351.3688750099391,
"inferenceMs": 379.56937501439825,
"imagesPerSecond": 2.634564498155487,
"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,
"modelInferenceMs": 343.3430829900317,
"totalMs": 2311.7804159992374,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3830956032,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": null,
"top1": "a person",
"logOddsMargin": null,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1932.2110409848392,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 346.9537920318544,
"inferenceMs": 348.4246250009164,
"imagesPerSecond": 2.8700612076352807,
"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,
"modelInferenceMs": 344.638540991582,
"totalMs": 2280.6356659857556,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3830972416,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "un chien",
"top1": "un chien",
"logOddsMargin": 9.700699366827896,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1932.2110409848392,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 362.0617500273511,
"inferenceMs": 404.293333995156,
"imagesPerSecond": 2.4734516152373205,
"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,
"modelInferenceMs": 351.02504200767726,
"totalMs": 2336.504374979995,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3833233408,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "一只狗",
"top1": "一只狗",
"logOddsMargin": 2.790970286157167,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1932.2110409848392,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 364.6842499729246,
"inferenceMs": 408.8544159894809,
"imagesPerSecond": 2.4458583811058268,
"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,
"modelInferenceMs": 352.9978330479935,
"totalMs": 2341.06545697432,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3837558784,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1932.2110409848392,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 363.1356250261888,
"inferenceMs": 407.4606249923818,
"imagesPerSecond": 2.4542248714675114,
"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,
"modelInferenceMs": 351.4027089695446,
"totalMs": 2339.671665977221,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3843276800,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1932.2110409848392,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 357.5835839728825,
"inferenceMs": 402.4658339913003,
"imagesPerSecond": 2.484682960744479,
"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,
"modelInferenceMs": 345.83004197338596,
"totalMs": 2334.6768749761395,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3848978432,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1932.2110409848392,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 355.59958301018924,
"inferenceMs": 401.40591602539644,
"imagesPerSecond": 2.4912438010423625,
"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,
"modelInferenceMs": 344.429292017594,
"totalMs": 2333.6169570102356,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3854680064,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1932.2110409848392,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 355.6690420373343,
"inferenceMs": 398.0750409536995,
"imagesPerSecond": 2.5120891719416063,
"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,
"modelInferenceMs": 344.51304195681587,
"totalMs": 2330.2860819385387,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3860381696,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1932.2110409848392,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 355.49341602018103,
"inferenceMs": 400.8286250173114,
"imagesPerSecond": 2.4948317998915646,
"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,
"modelInferenceMs": 344.36133300187066,
"totalMs": 2333.0396660021506,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3866116096,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1928.8584159803577,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 662.3090829816647,
"inferenceMs": 706.3075840123929,
"imagesPerSecond": 1.4158137653275629,
"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,
"modelInferenceMs": 643.8972500036471,
"totalMs": 2635.1659999927506,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3802578944,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1928.8584159803577,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 386.7258749669418,
"inferenceMs": 430.8383750030771,
"imagesPerSecond": 2.321056010836681,
"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,
"modelInferenceMs": 375.5751659628004,
"totalMs": 2359.6967909834348,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3808346112,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 10.95101885160621,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1928.8584159803577,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 357.8144999919459,
"inferenceMs": 401.44454198889434,
"imagesPerSecond": 2.491004099957758,
"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,
"modelInferenceMs": 346.76175005733967,
"totalMs": 2330.302957969252,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3815145472,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.524991987271818,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1928.8584159803577,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 356.66183300781995,
"inferenceMs": 392.67870801268145,
"imagesPerSecond": 2.5466112106279652,
"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,
"modelInferenceMs": 343.38995901634917,
"totalMs": 2321.537123993039,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3825188864,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.807835953516252,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1928.8584159803577,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 356.6129999817349,
"inferenceMs": 398.20079097989947,
"imagesPerSecond": 2.511295865433071,
"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,
"modelInferenceMs": 343.27625000150874,
"totalMs": 2327.059206960257,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3829628928,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.698087760556987,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1928.8584159803577,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 352.535167010501,
"inferenceMs": 382.9017499811016,
"imagesPerSecond": 2.611636013806037,
"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,
"modelInferenceMs": 344.534375006333,
"totalMs": 2311.7601659614593,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3829481472,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.76818182179674,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1928.8584159803577,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 352.0281660021283,
"inferenceMs": 382.7600000076927,
"imagesPerSecond": 2.612603197773806,
"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,
"modelInferenceMs": 343.9817909966223,
"totalMs": 2311.6184159880504,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3832561664,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": null,
"top1": "a person",
"logOddsMargin": null,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1928.8584159803577,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 346.54754103394225,
"inferenceMs": 348.03091699723154,
"imagesPerSecond": 2.8733079481210417,
"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,
"modelInferenceMs": 344.20329198474064,
"totalMs": 2276.889332977589,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3832578048,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "un chien",
"top1": "un chien",
"logOddsMargin": 9.700699366827896,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1928.8584159803577,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 355.2657079999335,
"inferenceMs": 398.5881669796072,
"imagesPerSecond": 2.508855211577725,
"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,
"modelInferenceMs": 344.2221249570139,
"totalMs": 2327.446582959965,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3834740736,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "一只狗",
"top1": "一只狗",
"logOddsMargin": 2.790970286157167,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1928.8584159803577,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 356.1268339981325,
"inferenceMs": 401.56787500018254,
"imagesPerSecond": 2.4902390411572277,
"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,
"modelInferenceMs": 345.0037910370156,
"totalMs": 2330.42629098054,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3839082496,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1928.8584159803577,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 355.8196250232868,
"inferenceMs": 401.8939579837024,
"imagesPerSecond": 2.4882185465464297,
"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,
"modelInferenceMs": 344.77091702865437,
"totalMs": 2330.75237396406,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3844751360,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1928.8584159803577,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 356.47016600705683,
"inferenceMs": 402.2380840033293,
"imagesPerSecond": 2.4860898054390175,
"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,
"modelInferenceMs": 345.4025409882888,
"totalMs": 2331.096499983687,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3850452992,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1928.8584159803577,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 355.0090420176275,
"inferenceMs": 400.30462498543784,
"imagesPerSecond": 2.498097542681096,
"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,
"modelInferenceMs": 343.81783299613744,
"totalMs": 2329.1630409657955,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3856171008,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1928.8584159803577,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 362.36229102360085,
"inferenceMs": 407.9957909998484,
"imagesPerSecond": 2.4510056771648694,
"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,
"modelInferenceMs": 351.0693749994971,
"totalMs": 2336.854206980206,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3861905408,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "cpu",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006509242482158,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 1928.8584159803577,
"placement": {
"cpu/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "PyTorch CPU",
"acceleration": "cpu",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 365.5667079729028,
"inferenceMs": 413.23137504514307,
"imagesPerSecond": 2.4199517761466103,
"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,
"modelInferenceMs": 353.7925000418909,
"totalMs": 2342.0897910255007,
"requestedDevice": "cpu",
"actualDevice": "cpu",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 3867607040,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2632.740625005681,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 273.71795900398865,
"inferenceMs": 319.03683400014415,
"imagesPerSecond": 3.134434314250837,
"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": 5589745664,
"modelInferenceMs": 241.19795800652355,
"totalMs": 2951.777459005825,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 494534656,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2632.740625005681,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 129.89358301274478,
"inferenceMs": 175.93670799396932,
"imagesPerSecond": 5.683862176358771,
"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": 5589745664,
"modelInferenceMs": 117.00245802057907,
"totalMs": 2808.6773329996504,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 497369088,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 10.951022176258594,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 5.4016709327697754e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2632.740625005681,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 126.24712497927248,
"inferenceMs": 171.3368339696899,
"imagesPerSecond": 5.8364566265821365,
"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": 5589745664,
"modelInferenceMs": 113.22833399754018,
"totalMs": 2804.077458975371,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 502988800,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.524990572425493,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 1.30385160446167e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2632.740625005681,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 131.0747090028599,
"inferenceMs": 167.6026249770075,
"imagesPerSecond": 5.9664936640293345,
"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": 5589745664,
"modelInferenceMs": 115.71691697463393,
"totalMs": 2800.3432499826886,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 520880128,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.807834057743548,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 1.1827796697616577e-7,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2632.740625005681,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 120.17404200742021,
"inferenceMs": 162.3581669991836,
"imagesPerSecond": 6.159222036579339,
"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": 5589745664,
"modelInferenceMs": 105.27025000192225,
"totalMs": 2795.0987920048647,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 526286848,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.698091432457698,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000004231929779052734,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2632.740625005681,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 116.59591598436236,
"inferenceMs": 147.48712501022965,
"imagesPerSecond": 6.780252852109229,
"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": 5589745664,
"modelInferenceMs": 106.79620795417577,
"totalMs": 2780.2277500159107,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 527433728,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.76815894415141,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 8.940696716308594e-7,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2632.740625005681,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 118.53624996729195,
"inferenceMs": 148.78754201345146,
"imagesPerSecond": 6.720992809395244,
"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": 5589745664,
"modelInferenceMs": 108.76962501788512,
"totalMs": 2781.5281670191325,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 528367616,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": null,
"top1": "a person",
"logOddsMargin": null,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 7.639755494892597e-10,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2632.740625005681,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 101.47987498203292,
"inferenceMs": 103.108000010252,
"imagesPerSecond": 9.698568490326359,
"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": 5589745664,
"modelInferenceMs": 97.01008297270164,
"totalMs": 2735.848625015933,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 528646144,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "un chien",
"top1": "un chien",
"logOddsMargin": 9.700717012504292,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 7.82310962677002e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2632.740625005681,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 122.17279098695144,
"inferenceMs": 168.3770000236109,
"imagesPerSecond": 5.939053432830931,
"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": 5589745664,
"modelInferenceMs": 108.93479100195691,
"totalMs": 2801.117625029292,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 531529728,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "一只狗",
"top1": "一只狗",
"logOddsMargin": 2.7909704311167927,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 8.381903171539307e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2632.740625005681,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 120.02574995858595,
"inferenceMs": 166.39279201626778,
"imagesPerSecond": 6.009875715663409,
"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": 5589745664,
"modelInferenceMs": 107.0600000093691,
"totalMs": 2799.133417021949,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 534315008,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2632.740625005681,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 123.7715000170283,
"inferenceMs": 170.02641700673848,
"imagesPerSecond": 5.881438999919454,
"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": 5589745664,
"modelInferenceMs": 111.37962498469278,
"totalMs": 2802.7670420124196,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 536100864,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2632.740625005681,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 122.98041698522866,
"inferenceMs": 166.6958329733461,
"imagesPerSecond": 5.998950196672855,
"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": 5589745664,
"modelInferenceMs": 110.7769170193933,
"totalMs": 2799.436457979027,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 538050560,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2632.740625005681,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 124.88600000506267,
"inferenceMs": 169.1681249649264,
"imagesPerSecond": 5.911279091184169,
"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": 5589745664,
"modelInferenceMs": 112.10329196183011,
"totalMs": 2801.9087499706075,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 543752192,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2632.740625005681,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 126.37066695606336,
"inferenceMs": 170.28704198310152,
"imagesPerSecond": 5.872437434782825,
"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": 5589745664,
"modelInferenceMs": 113.80133300554007,
"totalMs": 2803.0276669887826,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 549453824,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2632.740625005681,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 125.02112501533702,
"inferenceMs": 168.9994580228813,
"imagesPerSecond": 5.917178739499906,
"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": 5589745664,
"modelInferenceMs": 112.40483302390203,
"totalMs": 2801.7400830285624,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 555155456,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2658.178624988068,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 233.31949999555945,
"inferenceMs": 277.5425839936361,
"imagesPerSecond": 3.6030506944582226,
"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": 5589762048,
"modelInferenceMs": 203.8447079830803,
"totalMs": 2935.721208981704,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 491487232,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2658.178624988068,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 137.56200001807883,
"inferenceMs": 184.94191695936024,
"imagesPerSecond": 5.407103032352278,
"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": 5589745664,
"modelInferenceMs": 124.55929099814966,
"totalMs": 2843.120541947428,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 493109248,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 10.951022176258594,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 5.4016709327697754e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2658.178624988068,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 125.77333301305771,
"inferenceMs": 169.31212501367554,
"imagesPerSecond": 5.906251545299717,
"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": 5589745664,
"modelInferenceMs": 113.09595796046779,
"totalMs": 2827.4907500017434,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 498728960,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.524990572425493,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 1.30385160446167e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2658.178624988068,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 125.15166599769145,
"inferenceMs": 163.02454099059105,
"imagesPerSecond": 6.134045794109704,
"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": 5589745664,
"modelInferenceMs": 110.15083396341652,
"totalMs": 2821.203165978659,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 516620288,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.807834057743548,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 1.1827796697616577e-7,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2658.178624988068,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 130.03066700184718,
"inferenceMs": 169.93037500651553,
"imagesPerSecond": 5.8847630976019305,
"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": 5589745664,
"modelInferenceMs": 115.16254197340459,
"totalMs": 2828.1089999945834,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 522027008,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.698091432457698,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000004231929779052734,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2658.178624988068,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 116.29383399849758,
"inferenceMs": 147.6807079743594,
"imagesPerSecond": 6.771365154706747,
"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": 5589745664,
"modelInferenceMs": 106.26266605686396,
"totalMs": 2805.8593329624273,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 523157504,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.76815894415141,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 8.940696716308594e-7,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2658.178624988068,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 117.1377080027014,
"inferenceMs": 147.92066701920703,
"imagesPerSecond": 6.760380548244507,
"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": 5589745664,
"modelInferenceMs": 107.47149999951944,
"totalMs": 2806.099292007275,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 524091392,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": null,
"top1": "a person",
"logOddsMargin": null,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 7.639755494892597e-10,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2658.178624988068,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 100.0635419622995,
"inferenceMs": 101.43220901954919,
"imagesPerSecond": 9.85880135773508,
"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": 5589745664,
"modelInferenceMs": 95.59174999594688,
"totalMs": 2759.610834007617,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 524091392,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "un chien",
"top1": "un chien",
"logOddsMargin": 9.700717012504292,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 7.82310962677002e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2658.178624988068,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 122.0339160063304,
"inferenceMs": 166.44133400404826,
"imagesPerSecond": 6.008122958060872,
"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": 5589745664,
"modelInferenceMs": 108.60854102065787,
"totalMs": 2824.619958992116,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 526991360,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "一只狗",
"top1": "一只狗",
"logOddsMargin": 2.7909704311167927,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 8.381903171539307e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2658.178624988068,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 122.70974996499717,
"inferenceMs": 169.69666600925848,
"imagesPerSecond": 5.892867688664202,
"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": 5589745664,
"modelInferenceMs": 109.10354199586436,
"totalMs": 2827.8752909973264,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 529776640,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2658.178624988068,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 119.5936250151135,
"inferenceMs": 168.7712920247577,
"imagesPerSecond": 5.925178316779765,
"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": 5589745664,
"modelInferenceMs": 105.76554201543331,
"totalMs": 2826.9499170128256,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 533626880,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2658.178624988068,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 119.30612497963011,
"inferenceMs": 167.09333303151652,
"imagesPerSecond": 5.984679232003732,
"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": 5589745664,
"modelInferenceMs": 106.01362504530698,
"totalMs": 2825.2719580195844,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 535576576,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2658.178624988068,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 123.19762498373166,
"inferenceMs": 170.4028329695575,
"imagesPerSecond": 5.868447035611493,
"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": 5589745664,
"modelInferenceMs": 109.34737499337643,
"totalMs": 2828.5814579576254,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 541278208,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2658.178624988068,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 122.69616703270003,
"inferenceMs": 170.23104097461328,
"imagesPerSecond": 5.874369294076812,
"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": 5589745664,
"modelInferenceMs": 109.11120794480667,
"totalMs": 2828.409665962681,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 546979840,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2658.178624988068,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 121.90691701835021,
"inferenceMs": 169.1397909889929,
"imagesPerSecond": 5.91226933740906,
"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": 5589745664,
"modelInferenceMs": 108.05479204282165,
"totalMs": 2827.318415977061,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 552681472,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2654.433291987516,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 230.34312500385568,
"inferenceMs": 273.85024999966845,
"imagesPerSecond": 3.6516307726621053,
"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": 5589762048,
"modelInferenceMs": 201.6020420123823,
"totalMs": 2928.2835419871844,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 495222784,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2654.433291987516,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 121.06400000629947,
"inferenceMs": 165.73033400345594,
"imagesPerSecond": 6.033898417046256,
"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": 5589745664,
"modelInferenceMs": 108.00608299905434,
"totalMs": 2820.163625990972,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 496828416,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 10.951022176258594,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 5.4016709327697754e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2654.433291987516,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 122.93150002369657,
"inferenceMs": 165.78120802296326,
"imagesPerSecond": 6.032046767698089,
"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": 5589745664,
"modelInferenceMs": 110.36120797507465,
"totalMs": 2820.2145000104792,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 502448128,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.524990572425493,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 1.30385160446167e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2654.433291987516,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 124.412959034089,
"inferenceMs": 160.88337497785687,
"imagesPerSecond": 6.215682634315911,
"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": 5589745664,
"modelInferenceMs": 109.73254102282226,
"totalMs": 2815.316666965373,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 520339456,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.807834057743548,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 1.1827796697616577e-7,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2654.433291987516,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 121.48854095721617,
"inferenceMs": 162.8189580515027,
"imagesPerSecond": 6.141790931272765,
"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": 5589745664,
"modelInferenceMs": 106.78158298833296,
"totalMs": 2817.2522500390187,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 525746176,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.698091432457698,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000004231929779052734,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2654.433291987516,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 117.33704101061448,
"inferenceMs": 147.72120903944597,
"imagesPerSecond": 6.769508633881883,
"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": 5589745664,
"modelInferenceMs": 107.20274999039248,
"totalMs": 2802.154501026962,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 526876672,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.76815894415141,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 8.940696716308594e-7,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2654.433291987516,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 120.59795804088935,
"inferenceMs": 151.13458299310878,
"imagesPerSecond": 6.6166193084053875,
"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": 5589745664,
"modelInferenceMs": 110.56175001431257,
"totalMs": 2805.5678749806248,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 527810560,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": null,
"top1": "a person",
"logOddsMargin": null,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 7.639755494892597e-10,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2654.433291987516,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 99.27495801821351,
"inferenceMs": 100.67683400120586,
"imagesPerSecond": 9.93277162438404,
"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": 5589745664,
"modelInferenceMs": 95.45883402461186,
"totalMs": 2755.110125988722,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 527810560,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "un chien",
"top1": "un chien",
"logOddsMargin": 9.700717012504292,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 7.82310962677002e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2654.433291987516,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 123.4594999696128,
"inferenceMs": 169.6116660023108,
"imagesPerSecond": 5.895820868750714,
"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": 5589745664,
"modelInferenceMs": 109.64124999009073,
"totalMs": 2824.044957989827,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 530694144,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "一只狗",
"top1": "一只狗",
"logOddsMargin": 2.7909704311167927,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 8.381903171539307e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2654.433291987516,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 121.16325000533834,
"inferenceMs": 164.09404104342684,
"imagesPerSecond": 6.09406650991887,
"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": 5589745664,
"modelInferenceMs": 108.06791699724272,
"totalMs": 2818.527333030943,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 533479424,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2654.433291987516,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 123.96158400224522,
"inferenceMs": 169.77308300556615,
"imagesPerSecond": 5.890215234927519,
"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": 5589745664,
"modelInferenceMs": 111.82699998607859,
"totalMs": 2824.206374993082,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 535265280,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2654.433291987516,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 116.83308298233896,
"inferenceMs": 159.51495803892612,
"imagesPerSecond": 6.269004564173674,
"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": 5589745664,
"modelInferenceMs": 104.37249997630715,
"totalMs": 2813.948250026442,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 537214976,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2654.433291987516,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 119.24908304354176,
"inferenceMs": 165.8787499764003,
"imagesPerSecond": 6.028499733343004,
"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": 5589745664,
"modelInferenceMs": 106.68108297977597,
"totalMs": 2820.3120419639163,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 542916608,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2654.433291987516,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 122.92200000956655,
"inferenceMs": 168.41079201549292,
"imagesPerSecond": 5.937861748836174,
"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": 5589745664,
"modelInferenceMs": 109.84870896209031,
"totalMs": 2822.844084003009,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 548618240,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp32",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.006535026953568,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 2.421438694000244e-8,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2654.433291987516,
"placement": {
"mps:0/torch.float32": 888
},
"parameterDtype": "torch.float32",
"provider": "Metal/MPS",
"acceleration": "partial",
"accelerationDetails": "Image/text encoders and sigmoid logits on selected device; Pillow decode/EXIF/resize, tokenizer and JSON serialization on CPU",
"durationMs": 122.88904201705009,
"inferenceMs": 168.31112501677126,
"imagesPerSecond": 5.941377908919304,
"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": 5589745664,
"modelInferenceMs": 110.35574995912611,
"totalMs": 2822.7444170042872,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp32",
"actualPrecision": "fp32",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 554319872,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2329.351208987646,
"placement": {
"mps:0/torch.float16": 888
},
"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": 275.00470797531307,
"inferenceMs": 320.9042919916101,
"imagesPerSecond": 3.116193908762506,
"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": 2372730880,
"modelInferenceMs": 243.8192090485245,
"totalMs": 2650.255500979256,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 835436544,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2329.351208987646,
"placement": {
"mps:0/torch.float16": 888
},
"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": 116.61416699644178,
"inferenceMs": 162.16308303410187,
"imagesPerSecond": 6.166631648152041,
"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": 2372730880,
"modelInferenceMs": 102.97149996040389,
"totalMs": 2491.5142920217477,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 847806464,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 10.957031159204348,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.00016664713621139526,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2329.351208987646,
"placement": {
"mps:0/torch.float16": 888
},
"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": 117.77358298422769,
"inferenceMs": 162.83495799871162,
"imagesPerSecond": 6.141187447034022,
"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": 2372730880,
"modelInferenceMs": 105.20466696470976,
"totalMs": 2492.1861669863574,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 850968576,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.546875029154597,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.00011055544018745422,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2329.351208987646,
"placement": {
"mps:0/torch.float16": 888
},
"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": 110.66141596529633,
"inferenceMs": 147.76745799463242,
"imagesPerSecond": 6.767389881176169,
"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": 2372714496,
"modelInferenceMs": 95.8992499508895,
"totalMs": 2477.1186669822782,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 852934656,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.828124881494531,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000033278949558734894,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2329.351208987646,
"placement": {
"mps:0/torch.float16": 888
},
"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": 111.94304202217609,
"inferenceMs": 152.92529197176918,
"imagesPerSecond": 6.539140694821137,
"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": 2372714496,
"modelInferenceMs": 96.61458298796788,
"totalMs": 2482.276500959415,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 855441408,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.710937433877344,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.0005918145179748535,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2329.351208987646,
"placement": {
"mps:0/torch.float16": 888
},
"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": 107.64891694998369,
"inferenceMs": 139.28800000576302,
"imagesPerSecond": 7.179369363898004,
"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": 2372714496,
"modelInferenceMs": 97.84041700186208,
"totalMs": 2468.639208993409,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 856735744,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.781250162637669,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.00009149312973022461,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2329.351208987646,
"placement": {
"mps:0/torch.float16": 888
},
"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": 107.55062499083579,
"inferenceMs": 137.87995895836502,
"imagesPerSecond": 7.252685651741204,
"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": 2372714496,
"modelInferenceMs": 98.00566697958857,
"totalMs": 2467.231167946011,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 857849856,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": null,
"top1": "a person",
"logOddsMargin": null,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 4.491812433116138e-7,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2329.351208987646,
"placement": {
"mps:0/torch.float16": 888
},
"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": 82.52454100875184,
"inferenceMs": 83.96658301353455,
"imagesPerSecond": 11.9094997570499,
"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": 2372714496,
"modelInferenceMs": 78.50679097464308,
"totalMs": 2413.3177920011804,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 852557824,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "un chien",
"top1": "un chien",
"logOddsMargin": 9.69531245881724,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0.00006689317524433136,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2329.351208987646,
"placement": {
"mps:0/torch.float16": 888
},
"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": 108.04287501377985,
"inferenceMs": 152.952250035014,
"imagesPerSecond": 6.537988161475747,
"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": 2372714496,
"modelInferenceMs": 95.24037496885285,
"totalMs": 2482.30345902266,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 855687168,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "一只狗",
"top1": "一只狗",
"logOddsMargin": 2.7851561236948346,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0.00002637971192598343,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2329.351208987646,
"placement": {
"mps:0/torch.float16": 888
},
"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": 103.9633330074139,
"inferenceMs": 149.3541660020128,
"imagesPerSecond": 6.695494519962191,
"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": 2372714496,
"modelInferenceMs": 90.15275002457201,
"totalMs": 2478.7053749896586,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 858734592,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2329.351208987646,
"placement": {
"mps:0/torch.float16": 888
},
"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": 113.52533398894593,
"inferenceMs": 162.0345419505611,
"imagesPerSecond": 6.171523602079323,
"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": 2372714496,
"modelInferenceMs": 99.67179101658985,
"totalMs": 2491.385750938207,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 861978624,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2329.351208987646,
"placement": {
"mps:0/torch.float16": 888
},
"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": 106.08591604977846,
"inferenceMs": 156.2736249761656,
"imagesPerSecond": 6.399032467267059,
"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": 2372714496,
"modelInferenceMs": 92.58641599444672,
"totalMs": 2485.6248339638114,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 865058816,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2329.351208987646,
"placement": {
"mps:0/torch.float16": 888
},
"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": 108.80699998233467,
"inferenceMs": 155.6534589617513,
"imagesPerSecond": 6.42452796532925,
"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": 2372714496,
"modelInferenceMs": 94.3373329937458,
"totalMs": 2485.004667949397,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 868106240,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2329.351208987646,
"placement": {
"mps:0/torch.float16": 888
},
"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": 112.76383302174509,
"inferenceMs": 160.12562497053295,
"imagesPerSecond": 6.245096624503571,
"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": 2372714496,
"modelInferenceMs": 99.08187499968335,
"totalMs": 2489.4768339581788,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 871153664,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2329.351208987646,
"placement": {
"mps:0/torch.float16": 888
},
"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": 104.2035419959575,
"inferenceMs": 151.8467910354957,
"imagesPerSecond": 6.585585333615908,
"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": 2372714496,
"modelInferenceMs": 89.89308297168463,
"totalMs": 2481.1980000231415,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 874201088,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.7474169763736,
"placement": {
"mps:0/torch.float16": 888
},
"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": 290.99595797015354,
"inferenceMs": 334.62924999184906,
"imagesPerSecond": 2.988381918270319,
"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": 2372730880,
"modelInferenceMs": 261.71579101355746,
"totalMs": 2765.3766669682227,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 832520192,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.7474169763736,
"placement": {
"mps:0/torch.float16": 888
},
"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": 111.5977080189623,
"inferenceMs": 156.03058296255767,
"imagesPerSecond": 6.408999960219131,
"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": 2372730880,
"modelInferenceMs": 97.95349999330938,
"totalMs": 2586.7779999389313,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 844038144,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 10.957031159204348,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.00016664713621139526,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.7474169763736,
"placement": {
"mps:0/torch.float16": 888
},
"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": 117.4419159651734,
"inferenceMs": 160.3472499991767,
"imagesPerSecond": 6.236464922255507,
"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": 2372730880,
"modelInferenceMs": 105.01212498638779,
"totalMs": 2591.0946669755504,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 847233024,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.546875029154597,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.00011055544018745422,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.7474169763736,
"placement": {
"mps:0/torch.float16": 888
},
"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": 120.87545800022781,
"inferenceMs": 156.35025000665337,
"imagesPerSecond": 6.395896392602159,
"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": 2372714496,
"modelInferenceMs": 106.5254170098342,
"totalMs": 2587.097666983027,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 849231872,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.828124881494531,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000033278949558734894,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.7474169763736,
"placement": {
"mps:0/torch.float16": 888
},
"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": 111.3231250201352,
"inferenceMs": 150.3462079563178,
"imagesPerSecond": 6.651315078665263,
"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": 2372714496,
"modelInferenceMs": 96.05454100528732,
"totalMs": 2581.0936249326915,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 851738624,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.710937433877344,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.0005918145179748535,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.7474169763736,
"placement": {
"mps:0/torch.float16": 888
},
"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": 102.95475000748411,
"inferenceMs": 134.8133329884149,
"imagesPerSecond": 7.417663949350873,
"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": 2372714496,
"modelInferenceMs": 93.16008299356326,
"totalMs": 2565.5607499647886,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 853000192,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.781250162637669,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.00009149312973022461,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.7474169763736,
"placement": {
"mps:0/torch.float16": 888
},
"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": 98.46937499241903,
"inferenceMs": 127.91270902380347,
"imagesPerSecond": 7.817831454213892,
"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": 2372714496,
"modelInferenceMs": 89.31779197882861,
"totalMs": 2558.660126000177,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 854114304,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": null,
"top1": "a person",
"logOddsMargin": null,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 4.491812433116138e-7,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.7474169763736,
"placement": {
"mps:0/torch.float16": 888
},
"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": 82.01441605342552,
"inferenceMs": 83.45087501220405,
"imagesPerSecond": 11.983097838743545,
"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": 2372714496,
"modelInferenceMs": 78.34916602587327,
"totalMs": 2514.1982919885777,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 848838656,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "un chien",
"top1": "un chien",
"logOddsMargin": 9.69531245881724,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0.00006689317524433136,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.7474169763736,
"placement": {
"mps:0/torch.float16": 888
},
"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": 110.07804196560755,
"inferenceMs": 153.46324996789917,
"imagesPerSecond": 6.516218053567718,
"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": 2372714496,
"modelInferenceMs": 97.34991699224338,
"totalMs": 2584.210666944273,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 851902464,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "一只狗",
"top1": "一只狗",
"logOddsMargin": 2.7851561236948346,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0.00002637971192598343,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.7474169763736,
"placement": {
"mps:0/torch.float16": 888
},
"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": 113.22137503884733,
"inferenceMs": 159.29641597904265,
"imagesPerSecond": 6.277605141672252,
"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": 2372714496,
"modelInferenceMs": 100.6762080360204,
"totalMs": 2590.0438329554163,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 854966272,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.7474169763736,
"placement": {
"mps:0/torch.float16": 888
},
"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": 110.52920803194866,
"inferenceMs": 154.07487499760464,
"imagesPerSecond": 6.490350876582225,
"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": 2372714496,
"modelInferenceMs": 97.58841601433232,
"totalMs": 2584.8222919739783,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 858013696,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.7474169763736,
"placement": {
"mps:0/torch.float16": 888
},
"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": 106.4555419725366,
"inferenceMs": 152.3407920030877,
"imagesPerSecond": 6.564230019099098,
"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": 2372714496,
"modelInferenceMs": 94.16675003012642,
"totalMs": 2583.0882089794613,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 861077504,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.7474169763736,
"placement": {
"mps:0/torch.float16": 888
},
"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": 103.31216600025073,
"inferenceMs": 146.47687500109896,
"imagesPerSecond": 6.827016209845393,
"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": 2372714496,
"modelInferenceMs": 90.39412502897903,
"totalMs": 2577.2242919774726,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 864108544,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.7474169763736,
"placement": {
"mps:0/torch.float16": 888
},
"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": 107.64012503204867,
"inferenceMs": 153.45037501538172,
"imagesPerSecond": 6.516764784053222,
"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": 2372714496,
"modelInferenceMs": 94.87483301199973,
"totalMs": 2584.1977919917554,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 867188736,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2430.7474169763736,
"placement": {
"mps:0/torch.float16": 888
},
"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": 103.17524999845773,
"inferenceMs": 146.37241599848494,
"imagesPerSecond": 6.83188832525898,
"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": 2372714496,
"modelInferenceMs": 90.2908340212889,
"totalMs": 2577.1198329748586,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 870219776,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2371.0757910157554,
"placement": {
"mps:0/torch.float16": 888
},
"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": 235.917458019685,
"inferenceMs": 280.31399997416884,
"imagesPerSecond": 3.567427956121174,
"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": 2372730880,
"modelInferenceMs": 207.36245898297057,
"totalMs": 2651.389790989924,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 836141056,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "good",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2371.0757910157554,
"placement": {
"mps:0/torch.float16": 888
},
"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": 115.14487501699477,
"inferenceMs": 160.26099998271093,
"imagesPerSecond": 6.239821292191367,
"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": 2372730880,
"modelInferenceMs": 102.4924999801442,
"totalMs": 2531.3367909984663,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 847200256,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 10.957031159204348,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.00016664713621139526,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2371.0757910157554,
"placement": {
"mps:0/torch.float16": 888
},
"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": 118.55929100420326,
"inferenceMs": 162.33449999708682,
"imagesPerSecond": 6.1601199992481295,
"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": 2372730880,
"modelInferenceMs": 106.09404201386496,
"totalMs": 2533.410291012842,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 851099648,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.546875029154597,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.00011055544018745422,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2371.0757910157554,
"placement": {
"mps:0/torch.float16": 888
},
"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": 118.22416697395965,
"inferenceMs": 154.2222499847412,
"imagesPerSecond": 6.484148688655108,
"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": 2372714496,
"modelInferenceMs": 103.40370901394635,
"totalMs": 2525.2980410004966,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 853082112,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a person",
"top1": "a person",
"logOddsMargin": 9.828124881494531,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000033278949558734894,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2371.0757910157554,
"placement": {
"mps:0/torch.float16": 888
},
"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": 110.44404201675206,
"inferenceMs": 152.60487498017028,
"imagesPerSecond": 6.552870608687577,
"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": 2372714496,
"modelInferenceMs": 95.61474999645725,
"totalMs": 2523.6806659959257,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 855588864,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.710937433877344,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.0005918145179748535,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2371.0757910157554,
"placement": {
"mps:0/torch.float16": 888
},
"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": 103.69141603587195,
"inferenceMs": 134.2210000148043,
"imagesPerSecond": 7.450398968043019,
"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": 2372714496,
"modelInferenceMs": 93.84562500054017,
"totalMs": 2505.2967910305597,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 856899584,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a red car",
"top1": "a red car",
"logOddsMargin": 14.781250162637669,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.00009149312973022461,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2371.0757910157554,
"placement": {
"mps:0/torch.float16": 888
},
"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": 106.31666699191555,
"inferenceMs": 136.7008329834789,
"imagesPerSecond": 7.315244378363487,
"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": 2372714496,
"modelInferenceMs": 96.10541601432487,
"totalMs": 2507.7766239992343,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 858013696,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": null,
"top1": "a person",
"logOddsMargin": null,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 4.491812433116138e-7,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2371.0757910157554,
"placement": {
"mps:0/torch.float16": 888
},
"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": 81.60254103131592,
"inferenceMs": 83.02874997025356,
"imagesPerSecond": 12.044020900691228,
"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": 2372714496,
"modelInferenceMs": 77.90833304170519,
"totalMs": 2454.104540986009,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 852738048,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "un chien",
"top1": "un chien",
"logOddsMargin": 9.69531245881724,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0.00006689317524433136,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2371.0757910157554,
"placement": {
"mps:0/torch.float16": 888
},
"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": 112.86454199580476,
"inferenceMs": 158.206666004844,
"imagesPerSecond": 6.320846177046558,
"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": 2372714496,
"modelInferenceMs": 99.93224998470396,
"totalMs": 2529.2824570205994,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 855801856,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "一只狗",
"top1": "一只狗",
"logOddsMargin": 2.7851561236948346,
"minimumLogOddsMargin": 0.5,
"maximumCpuScoreError": 0.00002637971192598343,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2371.0757910157554,
"placement": {
"mps:0/torch.float16": 888
},
"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": 104.41149998223409,
"inferenceMs": 150.22029198007658,
"imagesPerSecond": 6.656890269742173,
"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": 2372714496,
"modelInferenceMs": 91.48229204583913,
"totalMs": 2521.296082995832,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 858849280,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2371.0757910157554,
"placement": {
"mps:0/torch.float16": 888
},
"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": 105.78954097582027,
"inferenceMs": 149.39404104370624,
"imagesPerSecond": 6.693707413051658,
"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": 2372714496,
"modelInferenceMs": 93.18712499225512,
"totalMs": 2520.4698320594616,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 861896704,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2371.0757910157554,
"placement": {
"mps:0/torch.float16": 888
},
"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": 110.09670799830928,
"inferenceMs": 153.90708402264863,
"imagesPerSecond": 6.49742671918105,
"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": 2372714496,
"modelInferenceMs": 97.23512502387166,
"totalMs": 2524.982875038404,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 864960512,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2371.0757910157554,
"placement": {
"mps:0/torch.float16": 888
},
"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": 103.01299998536706,
"inferenceMs": 146.66199998464435,
"imagesPerSecond": 6.818398767947396,
"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": 2372714496,
"modelInferenceMs": 90.29916604049504,
"totalMs": 2517.7377910003997,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 868007936,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2371.0757910157554,
"placement": {
"mps:0/torch.float16": 888
},
"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": 104.90520903840661,
"inferenceMs": 151.09795797616243,
"imagesPerSecond": 6.618223127527391,
"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": 2372714496,
"modelInferenceMs": 92.15358301298693,
"totalMs": 2522.173748991918,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 871071744,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
},
{
"precision": "fp16",
"device": "mps",
"status": "pass",
"quality": {
"passed": true,
"expectedLabel": "a dog",
"top1": "a dog",
"logOddsMargin": 11.000000066531332,
"minimumLogOddsMargin": 1,
"maximumCpuScoreError": 0.000016452744603157043,
"maximumAllowedCpuScoreError": 0.03,
"speedPassed": true,
"metric": "Known-object top1/log-odds separation and independent same-checkpointCPU score consistency; blank finite shape only"
},
"runtime": {
"engine": "transformers",
"version": "5.22.1",
"torchVersion": "2.10.0",
"modelType": "siglip",
"patchSize": 16,
"processorType": "SiglipImageProcessorPil",
"loadMs": 2371.0757910157554,
"placement": {
"mps:0/torch.float16": 888
},
"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": 102.79741697013378,
"inferenceMs": 148.1660830322653,
"imagesPerSecond": 6.749182940755986,
"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": 2372714496,
"modelInferenceMs": 89.3110420438461,
"totalMs": 2519.2418740480207,
"requestedDevice": "mps",
"actualDevice": "mps",
"requestedPrecision": "fp16",
"actualPrecision": "fp16",
"fallback": false,
"cpuThreads": 4,
"residentMemoryBytes": 874102784,
"memoryMeasurementScope": "After-call RSS and Metal driver allocation, not peak or validated minimum",
"grade": "excellent",
"artifactRevision": "365d321c0cfdea96bc28e3a29787a11a062681a1",
"helperSourceSha256": "2c7af3b327f015a16f1959c0c1e74b719cea475c0b63c3d29c9150967e4d4704"
}
}
]Release validation still required
- Public sample assets, complete dependency backup and forced original model/runtime source recovery excluded/pending