Download serving.json from vector-institute/mars-gend: direct link, hf CLI and curl.
- Browser
- Download file 1.02 kB
-
https://huggingface.co/vector-institute/mars-gend/resolve/main/serving.json
- Command line
-
hf download hf://vector-institute/mars-gend/serving.json
-
curl -L -o serving.json https://huggingface.co/vector-institute/mars-gend/resolve/main/serving.json
1.02 kB
| { | |
| "onnxruntime": "onnxruntime-gpu 1.32.0", | |
| "measured_on": "NVIDIA A100-SXM4-80GB, CUDA execution provider, arena_extend_strategy kSameAsRequested, cudnn_conv_algo_search HEURISTIC; peak is process total in GiB including the CUDA context; time is the median of 5 runs after 2 warm-up runs", | |
| "onnx_opset": 17, | |
| "batch_axis": "dynamic", | |
| "n_frames": 64, | |
| "encoder_chunk": 16, | |
| "outputs": [ | |
| "fake_prob", | |
| "logits", | |
| "frame_probs" | |
| ], | |
| "files": { | |
| "model.onnx": { | |
| "precision": "float32", | |
| "peak_vram_gb": { | |
| "batch1": 2.82, | |
| "batch2": 3.96 | |
| }, | |
| "seconds_per_clip": { | |
| "batch1": 0.59, | |
| "batch2": 0.57 | |
| } | |
| }, | |
| "model.fp16.onnx": { | |
| "precision": "float16", | |
| "peak_vram_gb": { | |
| "batch1": 2.25, | |
| "batch2": 3.43 | |
| }, | |
| "seconds_per_clip": { | |
| "batch1": 0.47, | |
| "batch2": 0.45 | |
| } | |
| } | |
| }, | |
| "pytorch_bf16_reference": { | |
| "precision": "bfloat16 autocast", | |
| "peak_vram_gb": { | |
| "batch1": 3.72, | |
| "batch2": 3.88 | |
| }, | |
| "seconds_per_clip": { | |
| "batch1": 0.22, | |
| "batch2": 0.22 | |
| } | |
| } | |
| } | |