{ "device": "NVIDIA GB10", "torch": "2.9.1+cu130", "workload": { "resolution": [ 1344, 768 ], "frames": 124, "steps": 12, "sampler_step": 1, "seed": 440420, "text_tokens": 100, "tokens": 37810, "hidden_shape": [ 37810, 5376 ], "rotation_shape": [ 1, 37810, 1, 48, 2, 2 ] }, "rounds": 8, "blocks": [ { "block": 0, "candidate_vs_baseline": { "equal": true, "different_elements": 0, "max_abs": 0.0, "mean_abs": 0.0 }, "baseline_vs_capture": { "equal": true, "different_elements": 0, "max_abs": 0.0, "mean_abs": 0.0 }, "candidate_vs_capture": { "equal": true, "different_elements": 0, "max_abs": 0.0, "mean_abs": 0.0 }, "baseline": { "timing": { "count": 8, "mean_s": 0.5090636406239355, "p50_s": 0.5093254634994082, "p90_s": 0.5097644505149219, "p95_s": 0.5097685157816159, "p99_s": 0.509771767994971, "min_s": 0.50723756296793, "max_s": 0.5097725810483098 }, "memory": { "allocated_bytes_before": 13656331264, "reserved_bytes_before": 23125295104, "peak_allocated_bytes": 18806702592, "peak_reserved_bytes": 23125295104, "peak_allocated_delta_bytes": 5150371328, "peak_reserved_delta_bytes": 0 }, "launches": { "profiled_iterations": 1, "runtime_kernel_launches": 0, "runtime_kernel_launches_per_block": 0.0, "positive_self_device_allocated_bytes": 14785398272, "device_kernel_launches": 0 } }, "candidate": { "timing": { "count": 8, "mean_s": 0.5060154675156809, "p50_s": 0.5055866409966256, "p90_s": 0.5077965697273612, "p95_s": 0.5078650188806932, "p99_s": 0.5079197782033589, "min_s": 0.5044256669934839, "max_s": 0.5079334680340253 }, "memory": { "allocated_bytes_before": 13656331264, "reserved_bytes_before": 23125295104, "peak_allocated_bytes": 18806702592, "peak_reserved_bytes": 23125295104, "peak_allocated_delta_bytes": 5150371328, "peak_reserved_delta_bytes": 0 }, "launches": { "profiled_iterations": 1, "runtime_kernel_launches": 86, "runtime_kernel_launches_per_block": 86.0, "positive_self_device_allocated_bytes": 14785398272, "device_kernel_launches": 99 } }, "p50_improvement_percent": 0.7340733520555491 }, { "block": 24, "candidate_vs_baseline": { "equal": true, "different_elements": 0, "max_abs": 0.0, "mean_abs": 0.0 }, "baseline_vs_capture": { "equal": true, "different_elements": 0, "max_abs": 0.0, "mean_abs": 0.0 }, "candidate_vs_capture": { "equal": true, "different_elements": 0, "max_abs": 0.0, "mean_abs": 0.0 }, "baseline": { "timing": { "count": 8, "mean_s": 0.5123310216222308, "p50_s": 0.5125731414882466, "p90_s": 0.5134933520865161, "p95_s": 0.5140004205662989, "p99_s": 0.5144060753501253, "min_s": 0.5090091329766437, "max_s": 0.5145074890460819 }, "memory": { "allocated_bytes_before": 13656331264, "reserved_bytes_before": 23125295104, "peak_allocated_bytes": 18806702592, "peak_reserved_bytes": 23125295104, "peak_allocated_delta_bytes": 5150371328, "peak_reserved_delta_bytes": 0 }, "launches": { "profiled_iterations": 1, "runtime_kernel_launches": 87, "runtime_kernel_launches_per_block": 87.0, "positive_self_device_allocated_bytes": 14785398272, "device_kernel_launches": 100 } }, "candidate": { "timing": { "count": 8, "mean_s": 0.5083635471310117, "p50_s": 0.5081502740213182, "p90_s": 0.5093831970880274, "p95_s": 0.5095810055354377, "p99_s": 0.509739252293366, "min_s": 0.5069579069968313, "max_s": 0.5097788139828481 }, "memory": { "allocated_bytes_before": 13656331264, "reserved_bytes_before": 23125295104, "peak_allocated_bytes": 18806702592, "peak_reserved_bytes": 23125295104, "peak_allocated_delta_bytes": 5150371328, "peak_reserved_delta_bytes": 0 }, "launches": { "profiled_iterations": 1, "runtime_kernel_launches": 86, "runtime_kernel_launches_per_block": 86.0, "positive_self_device_allocated_bytes": 14785398272, "device_kernel_launches": 99 } }, "p50_improvement_percent": 0.8628753847863981 }, { "block": 49, "candidate_vs_baseline": { "equal": true, "different_elements": 0, "max_abs": 0.0, "mean_abs": 0.0 }, "baseline_vs_capture": { "equal": true, "different_elements": 0, "max_abs": 0.0, "mean_abs": 0.0 }, "candidate_vs_capture": { "equal": true, "different_elements": 0, "max_abs": 0.0, "mean_abs": 0.0 }, "baseline": { "timing": { "count": 8, "mean_s": 0.5096271816291846, "p50_s": 0.5093877254694235, "p90_s": 0.5109008273051586, "p95_s": 0.5112217181624146, "p99_s": 0.5114784308482194, "min_s": 0.5076214010477997, "max_s": 0.5115426090196706 }, "memory": { "allocated_bytes_before": 13656331264, "reserved_bytes_before": 23125295104, "peak_allocated_bytes": 18806702592, "peak_reserved_bytes": 23125295104, "peak_allocated_delta_bytes": 5150371328, "peak_reserved_delta_bytes": 0 }, "launches": { "profiled_iterations": 1, "runtime_kernel_launches": 87, "runtime_kernel_launches_per_block": 87.0, "positive_self_device_allocated_bytes": 14785398272, "device_kernel_launches": 100 } }, "candidate": { "timing": { "count": 8, "mean_s": 0.506856961757876, "p50_s": 0.5066641345038079, "p90_s": 0.5086580028233585, "p95_s": 0.509256707908935, "p99_s": 0.5097356719773961, "min_s": 0.5044357180013321, "max_s": 0.5098554129945114 }, "memory": { "allocated_bytes_before": 13656331264, "reserved_bytes_before": 23125295104, "peak_allocated_bytes": 18806702592, "peak_reserved_bytes": 23125295104, "peak_allocated_delta_bytes": 5150371328, "peak_reserved_delta_bytes": 0 }, "launches": { "profiled_iterations": 1, "runtime_kernel_launches": 86, "runtime_kernel_launches_per_block": 86.0, "positive_self_device_allocated_bytes": 14785398272, "device_kernel_launches": 99 } }, "p50_improvement_percent": 0.5346793472704265 } ] }