| {"ts": "2025-12-19T19:54:31Z", "run": "ca1c99ac13bb4217b0fb0c501a249580", "impl": "gpt_oss_experts", "tags": {"family": "reference", "backend": "pytorch"}, "wl": {"name": "cuda_B1_S512_E2", "batch": 1, "seq_len": 512, "hidden_dim": 2880, "expert_dim": 5760, "num_experts": 2, "top_k": 2, "dtype": "float32", "device": "cuda"}, "env": {"torch": "2.8.0+cu128", "cuda": "12.8", "gpu": "NVIDIA L40S", "sm": "8.9", "py": "3.11.14", "plat": "Linux-6.12.58-82.121.amzn2023.x86_64-x86_64-with-glibc2.35"}, "lat_ms": {"p10": 2.591275999975551, "p50": 2.6265569999850413, "p90": 2.6390279999759514, "mean": 2.626043199961714, "iqr": 0.02082100013467425, "raw_times": [2.591275999975551, 2.6390279999759514, 2.6265569999850413, 2.618206999841277, 2.6551480000307492], "has_warnings": false, "reps": 5, "warmup": 2}, "compile_ms": 2.6624880001691054, "peak_bytes": 311252992, "ok": true, "absmax": 1.0818243026733398e-05, "corr": {"ok": true, "rtol": 0.01, "atol": 0.01, "absmax": 1.0818243026733398e-05, "mae": 1.0733322142186807e-06, "mse": 1.9560496885423495e-12, "ref": "naive_moe"}, "err": null} | |
| {"ts": "2025-12-19T19:54:31Z", "run": "ca1c99ac13bb4217b0fb0c501a249580", "impl": "gpt_oss_experts", "tags": {"family": "reference", "backend": "pytorch"}, "wl": {"name": "cuda_B1_S512_E4", "batch": 1, "seq_len": 512, "hidden_dim": 2880, "expert_dim": 5760, "num_experts": 4, "top_k": 2, "dtype": "float32", "device": "cuda"}, "env": {"torch": "2.8.0+cu128", "cuda": "12.8", "gpu": "NVIDIA L40S", "sm": "8.9", "py": "3.11.14", "plat": "Linux-6.12.58-82.121.amzn2023.x86_64-x86_64-with-glibc2.35"}, "lat_ms": {"p10": 3.913345000000845, "p50": 3.932325000050696, "p90": 3.941766000025382, "mean": 3.9370316000258754, "iqr": 0.02511100001356681, "raw_times": [3.941766000025382, 3.913345000000845, 3.916655000011815, 3.981067000040639, 3.932325000050696], "has_warnings": false, "reps": 5, "warmup": 2}, "compile_ms": 3.937866000114809, "peak_bytes": 632822272, "ok": true, "absmax": 7.82310962677002e-06, "corr": {"ok": true, "rtol": 0.01, "atol": 0.01, "absmax": 7.82310962677002e-06, "mae": 5.576844728238939e-07, "mse": 5.436189692842319e-13, "ref": "naive_moe"}, "err": null} | |
| {"ts": "2025-12-19T19:54:32Z", "run": "ca1c99ac13bb4217b0fb0c501a249580", "impl": "gpt_oss_experts", "tags": {"family": "reference", "backend": "pytorch"}, "wl": {"name": "cuda_B1_S1024_E2", "batch": 1, "seq_len": 1024, "hidden_dim": 2880, "expert_dim": 5760, "num_experts": 2, "top_k": 2, "dtype": "float32", "device": "cuda"}, "env": {"torch": "2.8.0+cu128", "cuda": "12.8", "gpu": "NVIDIA L40S", "sm": "8.9", "py": "3.11.14", "plat": "Linux-6.12.58-82.121.amzn2023.x86_64-x86_64-with-glibc2.35"}, "lat_ms": {"p10": 3.80903300015234, "p50": 3.849652999861064, "p90": 3.853734000131226, "mean": 3.837069200062615, "iqr": 0.039670999967711396, "raw_times": [3.8140630001635145, 3.8588630000049307, 3.80903300015234, 3.853734000131226, 3.849652999861064], "has_warnings": false, "reps": 5, "warmup": 2}, "compile_ms": 3.788761999885537, "peak_bytes": 645417472, "ok": true, "absmax": 1.5497207641601562e-05, "corr": {"ok": true, "rtol": 0.01, "atol": 0.01, "absmax": 1.5497207641601562e-05, "mae": 1.1454358173068613e-06, "mse": 2.2412421311207575e-12, "ref": "naive_moe"}, "err": null} | |
| {"ts": "2025-12-19T19:54:34Z", "run": "ca1c99ac13bb4217b0fb0c501a249580", "impl": "gpt_oss_experts", "tags": {"family": "reference", "backend": "pytorch"}, "wl": {"name": "cuda_B1_S1024_E4", "batch": 1, "seq_len": 1024, "hidden_dim": 2880, "expert_dim": 5760, "num_experts": 4, "top_k": 2, "dtype": "float32", "device": "cuda"}, "env": {"torch": "2.8.0+cu128", "cuda": "12.8", "gpu": "NVIDIA L40S", "sm": "8.9", "py": "3.11.14", "plat": "Linux-6.12.58-82.121.amzn2023.x86_64-x86_64-with-glibc2.35"}, "lat_ms": {"p10": 5.2778859999307315, "p50": 5.308016000071802, "p90": 5.336937000038233, "mean": 5.31205640004373, "iqr": 0.038680999978168984, "raw_times": [5.339187000117818, 5.336937000038233, 5.298256000060064, 5.308016000071802, 5.2778859999307315], "has_warnings": false, "reps": 5, "warmup": 2}, "compile_ms": 5.26179400003457, "peak_bytes": 657099264, "ok": true, "absmax": 6.556510925292969e-06, "corr": {"ok": true, "rtol": 0.01, "atol": 0.01, "absmax": 6.556510925292969e-06, "mae": 4.852234951613354e-07, "mse": 4.015021550906467e-13, "ref": "naive_moe"}, "err": null} | |
| {"ts": "2025-12-19T19:54:36Z", "run": "ca1c99ac13bb4217b0fb0c501a249580", "impl": "gpt_oss_experts", "tags": {"family": "reference", "backend": "pytorch"}, "wl": {"name": "cuda_B4_S512_E2", "batch": 4, "seq_len": 512, "hidden_dim": 2880, "expert_dim": 5760, "num_experts": 2, "top_k": 2, "dtype": "float32", "device": "cuda"}, "env": {"torch": "2.8.0+cu128", "cuda": "12.8", "gpu": "NVIDIA L40S", "sm": "8.9", "py": "3.11.14", "plat": "Linux-6.12.58-82.121.amzn2023.x86_64-x86_64-with-glibc2.35"}, "lat_ms": {"p10": 6.679864000034286, "p50": 6.717303999948854, "p90": 6.729205000056027, "mean": 6.711754200023279, "iqr": 0.028612000050998176, "raw_times": [6.679864000034286, 6.717303999948854, 6.700593000005028, 6.7318050000722, 6.729205000056027], "has_warnings": false, "reps": 5, "warmup": 2}, "compile_ms": 6.593322000071566, "peak_bytes": 678357504, "ok": true, "absmax": 1.3589859008789062e-05, "corr": {"ok": true, "rtol": 0.01, "atol": 0.01, "absmax": 1.3589859008789062e-05, "mae": 1.1745952406272409e-06, "mse": 2.316181968442521e-12, "ref": "naive_moe"}, "err": null} | |
| {"ts": "2025-12-19T19:54:38Z", "run": "ca1c99ac13bb4217b0fb0c501a249580", "impl": "gpt_oss_experts", "tags": {"family": "reference", "backend": "pytorch"}, "wl": {"name": "cuda_B4_S512_E4", "batch": 4, "seq_len": 512, "hidden_dim": 2880, "expert_dim": 5760, "num_experts": 4, "top_k": 2, "dtype": "float32", "device": "cuda"}, "env": {"torch": "2.8.0+cu128", "cuda": "12.8", "gpu": "NVIDIA L40S", "sm": "8.9", "py": "3.11.14", "plat": "Linux-6.12.58-82.121.amzn2023.x86_64-x86_64-with-glibc2.35"}, "lat_ms": {"p10": 7.423924000022453, "p50": 7.518165999954363, "p90": 7.529216999955679, "mean": 7.5042842000129895, "iqr": 0.02257999994981219, "raw_times": [7.543477000126586, 7.529216999955679, 7.518165999954363, 7.506637000005867, 7.423924000022453], "has_warnings": false, "reps": 5, "warmup": 2}, "compile_ms": 7.323180999946999, "peak_bytes": 701983232, "ok": true, "absmax": 8.58306884765625e-06, "corr": {"ok": true, "rtol": 0.01, "atol": 0.01, "absmax": 8.58306884765625e-06, "mae": 5.268635732136318e-07, "mse": 4.753664909623589e-13, "ref": "naive_moe"}, "err": null} | |
| {"ts": "2025-12-19T19:54:42Z", "run": "ca1c99ac13bb4217b0fb0c501a249580", "impl": "gpt_oss_experts", "tags": {"family": "reference", "backend": "pytorch"}, "wl": {"name": "cuda_B4_S1024_E2", "batch": 4, "seq_len": 1024, "hidden_dim": 2880, "expert_dim": 5760, "num_experts": 2, "top_k": 2, "dtype": "float32", "device": "cuda"}, "env": {"torch": "2.8.0+cu128", "cuda": "12.8", "gpu": "NVIDIA L40S", "sm": "8.9", "py": "3.11.14", "plat": "Linux-6.12.58-82.121.amzn2023.x86_64-x86_64-with-glibc2.35"}, "lat_ms": {"p10": 13.163481999981741, "p50": 13.23755299995355, "p90": 13.251324000066234, "mean": 13.23588719997133, "iqr": 0.04864200013798836, "raw_times": [13.163481999981741, 13.202681999928245, 13.23755299995355, 13.32439499992688, 13.251324000066234], "has_warnings": false, "reps": 5, "warmup": 2}, "compile_ms": 13.090128999920125, "peak_bytes": 1012207616, "ok": true, "absmax": 1.71661376953125e-05, "corr": {"ok": true, "rtol": 0.01, "atol": 0.01, "absmax": 1.71661376953125e-05, "mae": 1.797086838450923e-06, "mse": 5.3811247992252564e-12, "ref": "naive_moe"}, "err": null} | |
| {"ts": "2025-12-19T19:54:46Z", "run": "ca1c99ac13bb4217b0fb0c501a249580", "impl": "gpt_oss_experts", "tags": {"family": "reference", "backend": "pytorch"}, "wl": {"name": "cuda_B4_S1024_E4", "batch": 4, "seq_len": 1024, "hidden_dim": 2880, "expert_dim": 5760, "num_experts": 4, "top_k": 2, "dtype": "float32", "device": "cuda"}, "env": {"torch": "2.8.0+cu128", "cuda": "12.8", "gpu": "NVIDIA L40S", "sm": "8.9", "py": "3.11.14", "plat": "Linux-6.12.58-82.121.amzn2023.x86_64-x86_64-with-glibc2.35"}, "lat_ms": {"p10": 13.26829100003124, "p50": 13.362623000148233, "p90": 13.40691399991556, "mean": 13.346813000043767, "iqr": 0.1288519999889104, "raw_times": [13.40691399991556, 13.418175000197152, 13.26829100003124, 13.27806199992665, 13.362623000148233], "has_warnings": false, "reps": 5, "warmup": 2}, "compile_ms": 12.873562000095262, "peak_bytes": 910968320, "ok": true, "absmax": 8.344650268554688e-06, "corr": {"ok": true, "rtol": 0.01, "atol": 0.01, "absmax": 8.344650268554688e-06, "mae": 5.471991357808292e-07, "mse": 5.06310813587485e-13, "ref": "naive_moe"}, "err": null} | |