{ "environment": { "model": "Qwen/Qwen2.5-1.5B-Instruct", "revision": "989aa7980e4cf806f80c7fef2b1adb7bc71aa306", "torch": "2.11.0", "transformers": "4.57.6", "gpu": "mps", "platform": "macOS-26.6.2-arm64-arm-64bit", "dtype": "torch.float16", "attention": "sdpa" }, "arguments": { "model": "Qwen/Qwen2.5-1.5B-Instruct", "presets": [ "fintech_fraud" ], "fields": [ 4 ], "steps": 4, "repeats": 8, "warmup": 2, "output": "mac_mps_continuation.json", "device": "mps", "dtype": "float16" }, "cases": [ { "preset": "fintech_fraud", "fields": 4, "prefix_tokens": 291, "suffix_tokens": 28, "batch_padded_tokens": 32, "correctness": { "max_abs_logit_error": 0.0390625, "mean_abs_logit_error": 0.005111373495310545, "full_vocab_argmax_agreement": 1.0, "candidate_decisions_equal": true, "candidate_disagreements": [], "candidate_collision_fields": [] }, "decode": { "batch": { "median_ms": 222.27393699722597, "min_ms": 218.0314580036793, "p90_ms": 225.18500000296626, "samples_ms": [ 218.92137500981335, 218.0314580036793, 218.31691700208467, 221.5709159936523, 222.97695800079964, 225.18500000296626, 224.5981669984758, 223.11737500422169 ] }, "tree": { "median_ms": 92.20895849284716, "min_ms": 91.14341699751094, "p90_ms": 93.54945798986591, "samples_ms": [ 91.56537499802653, 91.14341699751094, 92.22362498985603, 91.70979099872056, 92.8215000021737, 93.54945798986591, 92.19429199583828, 92.9221249971306 ] } }, "end_to_end": { "batch": { "median_ms": 363.8185415038606, "min_ms": 360.70520800421946, "p90_ms": 368.3792500087293, "samples_ms": [ 362.8430419921642, 364.4052500021644, 360.70520800421946, 366.33004099712707, 368.3792500087293, 361.49850000219885, 363.23183300555684, 364.8465419973945 ] }, "tree": { "median_ms": 236.11554150556913, "min_ms": 235.26033300731797, "p90_ms": 242.17433400917798, "samples_ms": [ 235.6036670098547, 236.22883300413378, 237.20962500374299, 235.26033300731797, 242.17433400917798, 236.00225000700448, 235.61708399211057, 237.31533299724106 ] } }, "memory": { "batch": { "driver_high_water_mib": 3372.140625, "extra_driver_mib": 204.0 }, "tree": { "driver_high_water_mib": 3196.140625, "extra_driver_mib": 40.0 } }, "decode_speedup": 2.410546010173927, "end_to_end_speedup": 1.5408496161836913, "decode_speedup_ci95": [ 2.370532995531195, 2.427722584229317 ], "end_to_end_speedup_ci95": [ 1.5291795474631322, 1.5495758989220285 ] } ] }