h3-blackwell-runtime/benchmarks/gb10-post-lora-revert-deployment-smoke.json
2026-08-25 20:30:22 +07:00

43 lines
1.1 KiB
JSON

{
"device": "NVIDIA GB10",
"deployment": {
"compose_project": "h3-blackwell-runtime",
"service": "h3-hot-runtime",
"image": "sha256:1d340e14cb6fc45ccfdbe63dde8db2a2b3aea94b493702c8a08e1f8d5b4f7b83",
"feature_flags": {
"H3_NVFP4_MODULATE_FUSION": "1",
"H3_NVFP4_SWIGLU_FUSION": "1"
},
"state": "running",
"health_ready": true
},
"startup": {
"warmup_request_seconds": 30.730418314109556,
"vortex_extension_precompiled": true
},
"resident_smoke": {
"output": "/output/h3-blackwell-runtime/post-lora-revert-smoke.mp4",
"width": 320,
"height": 192,
"frames": 22,
"steps": 1,
"seed": 440420,
"attention": "sage2",
"text_conditioned_seconds": 1.4340007840073667,
"sampled_seconds": 0.2167754319962114,
"vae_decoded_seconds": 0.2895864300080575,
"request_seconds": 2.3209229299682193,
"wall_seconds": 2.3224709409987554,
"media_validation": {
"codec": "h264",
"width": 320,
"height": 192,
"frames": 22,
"duration_seconds": 0.916667
}
},
"tests": {
"nvfp4_cuda_tests": "3 passed"
},
"result": "pass"
}