{ "activation_backends": { "triton_cuda": 300 }, "audio_tokens": 207, "capability": [ 12, 0 ], "cuda": "12.8", "cuda_after_load_bytes": 19996112896, "cuda_forward_ms": 114115.6328125, "cuda_peak_bytes": 49839283712, "effective_runtime_modes": { "native_packed_matmul": 300 }, "gpu": "NVIDIA RTX PRO 6000 Blackwell Server Edition", "h3_dtype_views": 50, "load_seconds": 0.7637474560178816, "move_to_cuda_seconds": 1.5632987627759576, "orbitquant_linear_count": 300, "orbitquant_linear_cuda_ms": 109206.42547607422, "orbitquant_shape_cuda_ms": { "14336x5376": { "calls": 50, "total_ms": 22013.64224243164 }, "5376x28672": { "calls": 50, "total_ms": 43292.70489501953 }, "5376x7168": { "calls": 150, "total_ms": 32823.50001525879 }, "7168x5376": { "calls": 50, "total_ms": 11076.578323364258 } }, "orbitquant_shapes": { "14336x5376": 50, "5376x28672": 50, "5376x7168": 150, "7168x5376": 50 }, "output_finite": [ true, true ], "output_shapes": [ [ 1, 10545, 96 ], [ 1, 207, 32 ] ], "runtime_mode": "native_packed_matmul", "slowest_orbitquant_linears": [ { "cuda_ms": 913.251953125, "name": "transformer_blocks.2.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 911.2760620117188, "name": "transformer_blocks.0.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 901.734375, "name": "transformer_blocks.1.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 901.5783081054688, "name": "transformer_blocks.3.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 868.2454833984375, "name": "transformer_blocks.4.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 865.6199951171875, "name": "transformer_blocks.5.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.7640380859375, "name": "transformer_blocks.6.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.7313232421875, "name": "transformer_blocks.7.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.56640625, "name": "transformer_blocks.37.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.19775390625, "name": "transformer_blocks.35.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.1362915039062, "name": "transformer_blocks.47.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.0994262695312, "name": "transformer_blocks.39.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.0728149414062, "name": "transformer_blocks.43.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.0718383789062, "name": "transformer_blocks.40.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.0676879882812, "name": "transformer_blocks.33.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.0646362304688, "name": "transformer_blocks.45.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.0635986328125, "name": "transformer_blocks.49.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.060546875, "name": "transformer_blocks.36.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.0595092773438, "name": "transformer_blocks.42.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.0595092773438, "name": "transformer_blocks.46.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.0513305664062, "name": "transformer_blocks.44.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.0328979492188, "name": "transformer_blocks.24.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.0308227539062, "name": "transformer_blocks.17.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.0308227539062, "name": "transformer_blocks.20.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.0298461914062, "name": "transformer_blocks.41.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.0267333984375, "name": "transformer_blocks.16.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.0216064453125, "name": "transformer_blocks.34.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.0206298828125, "name": "transformer_blocks.29.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.0175170898438, "name": "transformer_blocks.22.ff.net.0.proj", "shape": "5376x28672" }, { "cuda_ms": 862.0165405273438, "name": "transformer_blocks.21.ff.net.0.proj", "shape": "5376x28672" } ], "status": "pass", "text_tokens": 512, "torch": "2.8.0+cu128", "transformer": "/workspace/orbitquant-h3-perf/model/transformer", "transformer_block_cuda_ms": [ { "cuda_ms": 3487.91259765625, "name": "transformer_blocks.0" }, { "cuda_ms": 2355.49658203125, "name": "transformer_blocks.1" }, { "cuda_ms": 2366.400146484375, "name": "transformer_blocks.2" }, { "cuda_ms": 2352.244384765625, "name": "transformer_blocks.3" }, { "cuda_ms": 2286.18408203125, "name": "transformer_blocks.4" }, { "cuda_ms": 2254.422607421875, "name": "transformer_blocks.5" }, { "cuda_ms": 2250.595947265625, "name": "transformer_blocks.6" }, { "cuda_ms": 2248.635986328125, "name": "transformer_blocks.7" }, { "cuda_ms": 2246.948486328125, "name": "transformer_blocks.8" }, { "cuda_ms": 2246.48046875, "name": "transformer_blocks.9" }, { "cuda_ms": 2246.338134765625, "name": "transformer_blocks.10" }, { "cuda_ms": 2246.347412109375, "name": "transformer_blocks.11" }, { "cuda_ms": 2246.5205078125, "name": "transformer_blocks.12" }, { "cuda_ms": 2246.44677734375, "name": "transformer_blocks.13" }, { "cuda_ms": 2246.46630859375, "name": "transformer_blocks.14" }, { "cuda_ms": 2246.548095703125, "name": "transformer_blocks.15" }, { "cuda_ms": 2246.48046875, "name": "transformer_blocks.16" }, { "cuda_ms": 2246.6708984375, "name": "transformer_blocks.17" }, { "cuda_ms": 2246.313720703125, "name": "transformer_blocks.18" }, { "cuda_ms": 2246.57373046875, "name": "transformer_blocks.19" }, { "cuda_ms": 2246.4384765625, "name": "transformer_blocks.20" }, { "cuda_ms": 2246.5458984375, "name": "transformer_blocks.21" }, { "cuda_ms": 2246.38330078125, "name": "transformer_blocks.22" }, { "cuda_ms": 2246.49365234375, "name": "transformer_blocks.23" }, { "cuda_ms": 2246.5146484375, "name": "transformer_blocks.24" }, { "cuda_ms": 2246.240966796875, "name": "transformer_blocks.25" }, { "cuda_ms": 2246.17626953125, "name": "transformer_blocks.26" }, { "cuda_ms": 2246.2880859375, "name": "transformer_blocks.27" }, { "cuda_ms": 2246.201904296875, "name": "transformer_blocks.28" }, { "cuda_ms": 2246.2490234375, "name": "transformer_blocks.29" }, { "cuda_ms": 2246.255126953125, "name": "transformer_blocks.30" }, { "cuda_ms": 2246.263427734375, "name": "transformer_blocks.31" }, { "cuda_ms": 2246.54296875, "name": "transformer_blocks.32" }, { "cuda_ms": 2246.833740234375, "name": "transformer_blocks.33" }, { "cuda_ms": 2246.46826171875, "name": "transformer_blocks.34" }, { "cuda_ms": 2246.546630859375, "name": "transformer_blocks.35" }, { "cuda_ms": 2246.568359375, "name": "transformer_blocks.36" }, { "cuda_ms": 2247.191162109375, "name": "transformer_blocks.37" }, { "cuda_ms": 2246.4765625, "name": "transformer_blocks.38" }, { "cuda_ms": 2246.99853515625, "name": "transformer_blocks.39" }, { "cuda_ms": 2246.68359375, "name": "transformer_blocks.40" }, { "cuda_ms": 2246.66064453125, "name": "transformer_blocks.41" }, { "cuda_ms": 2246.555419921875, "name": "transformer_blocks.42" }, { "cuda_ms": 2246.562255859375, "name": "transformer_blocks.43" }, { "cuda_ms": 2246.51025390625, "name": "transformer_blocks.44" }, { "cuda_ms": 2246.63623046875, "name": "transformer_blocks.45" }, { "cuda_ms": 2246.504150390625, "name": "transformer_blocks.46" }, { "cuda_ms": 2246.570556640625, "name": "transformer_blocks.47" }, { "cuda_ms": 2246.426025390625, "name": "transformer_blocks.48" }, { "cuda_ms": 2246.447998046875, "name": "transformer_blocks.49" } ], "video_tokens": 10545, "wall_seconds": 114.1175869773142 }