h3-blackwell-runtime/benchmarks/gb10-sage-strided-nhd-blocks-1344x768-124f-seed440420.json

116 lines
3.1 KiB
JSON
Raw Normal View History

{
"device": "NVIDIA GB10",
"torch": "2.9.1+cu130",
"resolution": [
1344,
768
],
"frames": 124,
"packed_tokens": 37810,
"steps": 12,
"sampler_step": 1,
"seed": 440420,
"attention": "sage2",
"fused_elementwise": true,
"warmup": 3,
"iterations": 10,
"results": [
{
"block": 0,
"parity": {
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0,
"reference_checksum": -8279139.0,
"candidate_checksum": -8279139.0
},
"hnd_timing": {
"count": 10,
"mean_s": 0.48708497728803196,
"p50_s": 0.4870640740264207,
"p90_s": 0.48792775589972737,
"p95_s": 0.488274192425888,
"p99_s": 0.4885513416468166,
"min_s": 0.48618641699431464,
"max_s": 0.4886206289520487
},
"strided_nhd_timing": {
"count": 10,
"mean_s": 0.471230285911588,
"p50_s": 0.4714779225178063,
"p90_s": 0.4723133562132716,
"p95_s": 0.4731061626109294,
"p99_s": 0.47374040772905573,
"min_s": 0.4685067450045608,
"max_s": 0.4738989690085873
},
"p50_speedup": 1.0330580728475696,
"p50_latency_reduction": 0.0320002076518765
},
{
"block": 24,
"parity": {
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0,
"reference_checksum": 89071408.0,
"candidate_checksum": 89071408.0
},
"hnd_timing": {
"count": 10,
"mean_s": 0.4880072570987977,
"p50_s": 0.4879330809926614,
"p90_s": 0.4888307601737324,
"p95_s": 0.48945235056162345,
"p99_s": 0.48994962287193633,
"min_s": 0.4864739279728383,
"max_s": 0.4900739409495145
},
"strided_nhd_timing": {
"count": 10,
"mean_s": 0.47193070990615527,
"p50_s": 0.4716692080255598,
"p90_s": 0.47343356219353155,
"p95_s": 0.4738115225860383,
"p99_s": 0.47411389090004374,
"min_s": 0.47014945396222174,
"max_s": 0.47418948297854513
},
"p50_speedup": 1.034481523683056,
"p50_latency_reduction": 0.033332179351344804
},
{
"block": 49,
"parity": {
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0,
"reference_checksum": -40617144320.0,
"candidate_checksum": -40617144320.0
},
"hnd_timing": {
"count": 10,
"mean_s": 0.4874046443030238,
"p50_s": 0.48766444699140266,
"p90_s": 0.48841265672817824,
"p95_s": 0.48861716688843443,
"p99_s": 0.4887807750166394,
"min_s": 0.4857746610068716,
"max_s": 0.4888216770486906
},
"strided_nhd_timing": {
"count": 10,
"mean_s": 0.47163516000728123,
"p50_s": 0.47158038747147657,
"p90_s": 0.47319388142204843,
"p95_s": 0.4735081272228854,
"p99_s": 0.47375952386355497,
"min_s": 0.46926438697846606,
"max_s": 0.4738223730237223
},
"p50_speedup": 1.0341067184879458,
"p50_latency_reduction": 0.032981816942274755
}
]
}