File size: 872 Bytes
d2446c1 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 | {
"activation_dtype": "float32",
"all_matmul_tf32_qualified": false,
"autocast": false,
"cuda_matmul_allow_tf32": false,
"cudnn_allow_tf32": true,
"device": "cuda",
"id": "cuda-fp32-cudnn-tf32/1",
"input": "RGB640directresize",
"new_packaged_example_cuda_rerun": false,
"onnx_artifact_included": false,
"preprocessor_sha256": "cd38cd59999e7a95d68e487fbe5132df3d4e5c32a0836add57e6126ba0c4eaf1",
"qualification_source": "existing final-H5 six-panel CUDA evaluation and exact EMA export",
"schema": "ax-doclayout.runtime-profile/1",
"storage_dtype": "float32",
"tensorrt_qualified": false,
"tested_cuda": "12.9",
"tested_cudnn": 92000,
"tested_gpu": "NVIDIA RTX PRO 6000 Blackwell Server Edition",
"tested_torch": "2.12.1+cu129",
"tested_transformers": "4.53.3",
"tf32_scope": "cuDNN operations only; GEMM matmul TF32 is disabled"
}
|