melihcatal's picture
Add files using upload-large-folder tool
076fd74 verified
{"timestamp": 1774558080.5418942, "event": "train_step", "step": 5, "epoch": 1, "metrics": {"train/step_loss": 1.9771458762032645, "train/step_real_loss": 1.2914095222949982, "train/lr": 2.0320937499999996e-05, "train/step_canary_loss": 9.291666984558105, "perf/step_duration_sec": 4.801711035193875, "perf/samples_per_sec": 7.289068364062193, "perf/tokens_per_sec": 5446.391881627271, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 26152.0, "perf/physical_batches": 5.0, "privacy/epsilon": 0.2650253115825195, "system/cuda_memory_allocated_gb": 16.436955451965332, "system/cuda_max_memory_allocated_gb": 74.68954753875732}}
{"timestamp": 1774558106.1998873, "event": "train_step", "step": 10, "epoch": 1, "metrics": {"train/step_loss": 1.8103746245889103, "train/step_real_loss": 1.3317261636257172, "train/lr": 4.063187499999999e-05, "train/step_canary_loss": 9.46875, "perf/step_duration_sec": 5.370200600009412, "perf/samples_per_sec": 6.331234628356418, "perf/tokens_per_sec": 4692.9345618776015, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 25202.0, "perf/physical_batches": 5.0, "privacy/epsilon": 0.3609145567616601, "system/cuda_memory_allocated_gb": 16.342227458953857, "system/cuda_max_memory_allocated_gb": 74.68959474563599}}
{"timestamp": 1774558120.4473536, "event": "eval_step", "step": 10, "epoch": 1, "metrics": {"eval/loss": 0.8437587110833689, "eval/duration_sec": 14.230815590824932}}
{"timestamp": 1774558145.2432938, "event": "train_step", "step": 15, "epoch": 1, "metrics": {"train/step_loss": 1.9761825561523438, "train/step_real_loss": 1.3040277361869812, "train/lr": 6.094281249999999e-05, "train/step_canary_loss": 9.145833969116211, "perf/step_duration_sec": 4.77954536001198, "perf/samples_per_sec": 7.322872232331377, "perf/tokens_per_sec": 5590.280662161773, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 26719.0, "perf/physical_batches": 5.0, "privacy/epsilon": 0.4363307743601278, "system/cuda_memory_allocated_gb": 16.436955451965332, "system/cuda_max_memory_allocated_gb": 74.68959474563599}}
{"timestamp": 1774558171.2830229, "event": "train_step", "step": 20, "epoch": 1, "metrics": {"train/step_loss": 2.3287488884396024, "train/step_real_loss": 1.4694518744945526, "train/lr": 8.125375000000001e-05, "train/step_canary_loss": 9.203125, "perf/step_duration_sec": 5.488770961062983, "perf/samples_per_sec": 6.5588453690966295, "perf/tokens_per_sec": 4261.245398272252, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 23389.0, "perf/physical_batches": 5.0, "privacy/epsilon": 0.5009308261515277, "system/cuda_memory_allocated_gb": 16.531683444976807, "system/cuda_max_memory_allocated_gb": 74.68959474563599}}
{"timestamp": 1774558185.5570998, "event": "eval_step", "step": 20, "epoch": 1, "metrics": {"eval/loss": 0.8437518252229149, "eval/duration_sec": 14.25888997898437}}
{"timestamp": 1774558209.6112597, "event": "train_step", "step": 25, "epoch": 1, "metrics": {"train/step_loss": 2.2755661805470786, "train/step_real_loss": 1.3705588281154633, "train/lr": 0.00010156468750000002, "train/step_canary_loss": 9.515625, "perf/step_duration_sec": 4.738336017122492, "perf/samples_per_sec": 7.597603857115681, "perf/tokens_per_sec": 5560.390800650745, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 26347.0, "perf/physical_batches": 5.0, "privacy/epsilon": 0.5585751916186127, "system/cuda_memory_allocated_gb": 16.531683444976807, "system/cuda_max_memory_allocated_gb": 74.68959474563599}}
{"timestamp": 1774558235.9608197, "event": "train_step", "step": 30, "epoch": 1, "metrics": {"train/step_loss": 1.1493696885950424, "train/step_real_loss": 1.2212052941322327, "train/lr": 0.00012187562500000007, "train/step_canary_loss": 0.0, "perf/step_duration_sec": 5.743357056053355, "perf/samples_per_sec": 5.571654293419351, "perf/tokens_per_sec": 4864.054198155093, "perf/logical_batch_size": 32.0, "perf/logical_token_count": 27936.0, "perf/physical_batches": 6.0, "privacy/epsilon": 0.6109394245209149, "system/cuda_memory_allocated_gb": 16.909897327423096, "system/cuda_max_memory_allocated_gb": 74.68959474563599}}
{"timestamp": 1774558250.2758534, "event": "eval_step", "step": 30, "epoch": 1, "metrics": {"eval/loss": 0.843777087432417, "eval/duration_sec": 14.303893523989245}}
{"timestamp": 1774558274.64641, "event": "train_step", "step": 35, "epoch": 1, "metrics": {"train/step_loss": 1.7862528913161333, "train/step_real_loss": 1.2748468220233917, "train/lr": 0.00012999217477858396, "train/step_canary_loss": 9.96875, "perf/step_duration_sec": 4.786091385176405, "perf/samples_per_sec": 7.10391784521825, "perf/tokens_per_sec": 4982.771552140142, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 23848.0, "perf/physical_batches": 5.0, "privacy/epsilon": 0.6595864137332423, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68959474563599}}
{"timestamp": 1774558300.962021, "event": "train_step", "step": 40, "epoch": 1, "metrics": {"train/step_loss": 1.8911787762361414, "train/step_real_loss": 1.3746118247509003, "train/lr": 0.00012994436116313906, "train/step_canary_loss": 10.15625, "perf/step_duration_sec": 4.624199023004621, "perf/samples_per_sec": 7.352624709891519, "perf/tokens_per_sec": 5071.148513145768, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 23450.0, "perf/physical_batches": 5.0, "privacy/epsilon": 0.7050401828636351, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68964195251465}}
{"timestamp": 1774558315.2827964, "event": "eval_step", "step": 40, "epoch": 1, "metrics": {"eval/loss": 0.8438989198343319, "eval/duration_sec": 14.308534864801913}}
{"timestamp": 1774558339.5602052, "event": "train_step", "step": 45, "epoch": 1, "metrics": {"train/step_loss": 2.249728965759277, "train/step_real_loss": 1.3981409966945648, "train/lr": 0.0001298531148973165, "train/step_canary_loss": 11.333333969116211, "perf/step_duration_sec": 4.993837644113228, "perf/samples_per_sec": 7.008637944258812, "perf/tokens_per_sec": 5504.984735017685, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 27491.0, "perf/physical_batches": 5.0, "privacy/epsilon": 0.748051700589423, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68964195251465}}
{"timestamp": 1774558366.1590881, "event": "train_step", "step": 50, "epoch": 1, "metrics": {"train/step_loss": 1.5222526608091411, "train/step_real_loss": 1.2553699314594269, "train/lr": 0.0001297185002182214, "train/step_canary_loss": 10.0625, "perf/step_duration_sec": 5.099501257063821, "perf/samples_per_sec": 6.47122107368607, "perf/tokens_per_sec": 5423.471552573807, "perf/logical_batch_size": 33.0, "perf/logical_token_count": 27657.0, "perf/physical_batches": 5.0, "privacy/epsilon": 0.7890424569714427, "system/cuda_memory_allocated_gb": 16.247522354125977, "system/cuda_max_memory_allocated_gb": 74.68964195251465}}
{"timestamp": 1774558380.487505, "event": "eval_step", "step": 50, "epoch": 1, "metrics": {"eval/loss": 0.8441700407049872, "eval/duration_sec": 14.316868709865957}}
{"timestamp": 1774558405.4083753, "event": "train_step", "step": 55, "epoch": 1, "metrics": {"train/step_loss": 1.9350511823381698, "train/step_real_loss": 1.3059152960777283, "train/lr": 0.00012954061189419548, "train/step_canary_loss": 8.645833969116211, "perf/step_duration_sec": 5.407421349780634, "perf/samples_per_sec": 6.472586050913134, "perf/tokens_per_sec": 5246.123459780108, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 28368.0, "perf/physical_batches": 5.0, "privacy/epsilon": 0.8282486995209407, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68964195251465}}
{"timestamp": 1774558430.6544044, "event": "train_step", "step": 60, "epoch": 1, "metrics": {"train/step_loss": 2.754690647125244, "train/step_real_loss": 1.3903357684612274, "train/lr": 0.00012931957515810007, "train/step_canary_loss": 10.03125, "perf/step_duration_sec": 4.870435405056924, "perf/samples_per_sec": 7.802177185338499, "perf/tokens_per_sec": 5313.2826632155175, "perf/logical_batch_size": 38.0, "perf/logical_token_count": 25878.0, "perf/physical_batches": 5.0, "privacy/epsilon": 0.865459735721413, "system/cuda_memory_allocated_gb": 16.721160888671875, "system/cuda_max_memory_allocated_gb": 74.68964195251465}}
{"timestamp": 1774558444.9645379, "event": "eval_step", "step": 60, "epoch": 1, "metrics": {"eval/loss": 0.844405337152156, "eval/duration_sec": 14.298058781772852}}
{"timestamp": 1774558469.8114667, "event": "train_step", "step": 65, "epoch": 1, "metrics": {"train/step_loss": 2.289840751224094, "train/step_real_loss": 1.3436489701271057, "train/lr": 0.00012905554561915293, "train/step_canary_loss": 9.859375, "perf/step_duration_sec": 5.438727655913681, "perf/samples_per_sec": 6.619195201079097, "perf/tokens_per_sec": 5119.579754967955, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 27844.0, "perf/physical_batches": 5.0, "privacy/epsilon": 0.9017532578504478, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68964195251465}}
{"timestamp": 1774558495.208508, "event": "train_step", "step": 70, "epoch": 1, "metrics": {"train/step_loss": 2.236192994647556, "train/step_real_loss": 1.3604436814785004, "train/lr": 0.0001287487091533796, "train/step_canary_loss": 9.2421875, "perf/step_duration_sec": 4.876413284102455, "perf/samples_per_sec": 7.382475172349159, "perf/tokens_per_sec": 5855.533224201608, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 28554.0, "perf/physical_batches": 5.0, "privacy/epsilon": 0.9364996140118174, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68964195251465}}
{"timestamp": 1774558509.528138, "event": "eval_step", "step": 70, "epoch": 1, "metrics": {"eval/loss": 0.8446683649989692, "eval/duration_sec": 14.306719899876043}}
{"timestamp": 1774558535.405044, "event": "train_step", "step": 75, "epoch": 1, "metrics": {"train/step_loss": 2.5103939417246224, "train/step_real_loss": 1.2737367153167725, "train/lr": 0.0001283992817727574, "train/step_canary_loss": 10.425000190734863, "perf/step_duration_sec": 5.485939267091453, "perf/samples_per_sec": 6.74451505906239, "perf/tokens_per_sec": 4691.448218245642, "perf/logical_batch_size": 37.0, "perf/logical_token_count": 25737.0, "perf/physical_batches": 5.0, "privacy/epsilon": 0.9703952086694796, "system/cuda_memory_allocated_gb": 16.6264328956604, "system/cuda_max_memory_allocated_gb": 74.68964195251465}}
{"timestamp": 1774558561.1338398, "event": "train_step", "step": 80, "epoch": 1, "metrics": {"train/step_loss": 2.1728877226511636, "train/step_real_loss": 1.2179361879825592, "train/lr": 0.0001280075094731435, "train/step_canary_loss": 9.8125, "perf/step_duration_sec": 4.733840444125235, "perf/samples_per_sec": 7.604819052293265, "perf/tokens_per_sec": 5797.195812557891, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 27443.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.0029643968647324, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68964195251465}}
{"timestamp": 1774558575.596743, "event": "eval_step", "step": 80, "epoch": 1, "metrics": {"eval/loss": 0.8449255106124011, "eval/duration_sec": 14.451128582004458}}
{"timestamp": 1774558601.8340118, "event": "train_step", "step": 85, "epoch": 1, "metrics": {"train/step_loss": 1.983102798461914, "train/step_real_loss": 1.3687654733657837, "train/lr": 0.0001275736680610945, "train/step_canary_loss": 11.8125, "perf/step_duration_sec": 5.391784159000963, "perf/samples_per_sec": 6.305890406098864, "perf/tokens_per_sec": 5123.535954955326, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 27625.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.0353881587805573, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68964195251465}}
{"timestamp": 1774558626.859501, "event": "train_step", "step": 90, "epoch": 1, "metrics": {"train/step_loss": 2.0590727397373745, "train/step_real_loss": 1.3536732494831085, "train/lr": 0.00012709806295969954, "train/step_canary_loss": 9.583333969116211, "perf/step_duration_sec": 4.7415986829437315, "perf/samples_per_sec": 7.381476658052156, "perf/tokens_per_sec": 5354.1013690848595, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 25387.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.0658138978238574, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68964195251465}}
{"timestamp": 1774558641.166878, "event": "eval_step", "step": 90, "epoch": 1, "metrics": {"eval/loss": 0.8451848497444935, "eval/duration_sec": 14.289168786024675}}
{"timestamp": 1774558667.1133275, "event": "train_step", "step": 95, "epoch": 1, "metrics": {"train/step_loss": 2.289924965964423, "train/step_real_loss": 1.3808530867099762, "train/lr": 0.0001265810289935633, "train/step_canary_loss": 9.5625, "perf/step_duration_sec": 4.699590467847884, "perf/samples_per_sec": 7.660241939439827, "perf/tokens_per_sec": 5347.274442725636, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 25130.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.0962396368671574, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68964195251465}}
{"timestamp": 1774558691.82777, "event": "train_step", "step": 100, "epoch": 1, "metrics": {"train/step_loss": 2.768649468055138, "train/step_real_loss": 1.250268042087555, "train/lr": 0.00012602293015309042, "train/step_canary_loss": 9.709821701049805, "perf/step_duration_sec": 4.988021387020126, "perf/samples_per_sec": 7.818731511754571, "perf/tokens_per_sec": 6027.440074382082, "perf/logical_batch_size": 39.0, "perf/logical_token_count": 30065.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.126226314471096, "system/cuda_memory_allocated_gb": 16.81588888168335, "system/cuda_max_memory_allocated_gb": 74.68964195251465}}
{"timestamp": 1774558706.217189, "event": "eval_step", "step": 100, "epoch": 1, "metrics": {"eval/loss": 0.8454189227724617, "eval/duration_sec": 14.370606791926548}}
{"timestamp": 1774558732.984083, "event": "train_step", "step": 105, "epoch": 1, "metrics": {"train/step_loss": 2.2011702128819057, "train/step_real_loss": 1.378233015537262, "train/lr": 0.00012542415933823736, "train/step_canary_loss": 10.979166984558105, "perf/step_duration_sec": 5.005711684003472, "perf/samples_per_sec": 6.992012766505895, "perf/tokens_per_sec": 6105.425547712947, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 30562.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.1545412906837556, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774558758.1004992, "event": "train_step", "step": 110, "epoch": 1, "metrics": {"train/step_loss": 1.7460931609658634, "train/step_real_loss": 1.2263177335262299, "train/lr": 0.00012478513808191193, "train/step_canary_loss": 10.0625, "perf/step_duration_sec": 5.294084263034165, "perf/samples_per_sec": 6.4222627201845475, "perf/tokens_per_sec": 5047.520717905043, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 26722.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.1828562668964149, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774558772.4011295, "event": "eval_step", "step": 110, "epoch": 1, "metrics": {"eval/loss": 0.8455666301941328, "eval/duration_sec": 14.288897258928046}}
{"timestamp": 1774558798.8395534, "event": "train_step", "step": 115, "epoch": 1, "metrics": {"train/step_loss": 2.1378304072788783, "train/step_real_loss": 1.3948926329612732, "train/lr": 0.00012410631625321545, "train/step_canary_loss": 10.0625, "perf/step_duration_sec": 4.789301107171923, "perf/samples_per_sec": 7.307955632104213, "perf/tokens_per_sec": 4683.564365249417, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 22431.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.2111712431090742, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774558823.9987817, "event": "train_step", "step": 120, "epoch": 1, "metrics": {"train/step_loss": 2.0787572043282645, "train/step_real_loss": 1.338093787431717, "train/lr": 0.00012338817174073607, "train/step_canary_loss": 9.979166984558105, "perf/step_duration_sec": 5.2293452909216285, "perf/samples_per_sec": 6.692998464025989, "perf/tokens_per_sec": 5220.730030467816, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 27301.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.2381592500412848, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774558838.6896732, "event": "eval_step", "step": 120, "epoch": 1, "metrics": {"eval/loss": 0.8458415837111799, "eval/duration_sec": 14.674860873026773}}
{"timestamp": 1774558864.0763938, "event": "train_step", "step": 125, "epoch": 1, "metrics": {"train/step_loss": 2.9109930869860525, "train/step_real_loss": 1.3837102353572845, "train/lr": 0.00012263121011611664, "train/step_canary_loss": 9.892857551574707, "perf/step_duration_sec": 4.967141222208738, "perf/samples_per_sec": 7.851598787976049, "perf/tokens_per_sec": 5058.04020382334, "perf/logical_batch_size": 39.0, "perf/logical_token_count": 25124.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.2643951659168622, "system/cuda_memory_allocated_gb": 16.81588888168335, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774558889.6720726, "event": "train_step", "step": 130, "epoch": 1, "metrics": {"train/step_loss": 2.43642520904541, "train/step_real_loss": 1.2780540883541107, "train/lr": 0.00012183596427813386, "train/step_canary_loss": 9.850000381469727, "perf/step_duration_sec": 5.437113648978993, "perf/samples_per_sec": 6.80508122300295, "perf/tokens_per_sec": 4811.192424663086, "perf/logical_batch_size": 37.0, "perf/logical_token_count": 26159.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.2906310817924396, "system/cuda_memory_allocated_gb": 16.6264328956604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774558903.9675303, "event": "eval_step", "step": 130, "epoch": 1, "metrics": {"eval/loss": 0.8460407011549581, "eval/duration_sec": 14.284140298841521}}
{"timestamp": 1774558928.901615, "event": "train_step", "step": 135, "epoch": 1, "metrics": {"train/step_loss": 2.397670127250053, "train/step_real_loss": 1.3504310250282288, "train/lr": 0.00012100299407753897, "train/step_canary_loss": 9.100000381469727, "perf/step_duration_sec": 4.752518580993637, "perf/samples_per_sec": 7.785345679230188, "perf/tokens_per_sec": 5662.892115273569, "perf/logical_batch_size": 37.0, "perf/logical_token_count": 26913.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.316866997668017, "system/cuda_memory_allocated_gb": 16.6264328956604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774558955.0112307, "event": "train_step", "step": 140, "epoch": 1, "metrics": {"train/step_loss": 2.092343848092215, "train/step_real_loss": 1.3236573040485382, "train/lr": 0.00012013288592292455, "train/step_canary_loss": 10.291666984558105, "perf/step_duration_sec": 5.596101293107495, "perf/samples_per_sec": 6.254354266801455, "perf/tokens_per_sec": 4683.975258325707, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 26212.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.3428753209464988, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774558969.3190796, "event": "eval_step", "step": 140, "epoch": 1, "metrics": {"eval/loss": 0.8462591745298018, "eval/duration_sec": 14.296601020963863}}
{"timestamp": 1774558994.1577423, "event": "train_step", "step": 145, "epoch": 1, "metrics": {"train/step_loss": 1.6519569050181995, "train/step_real_loss": 1.3305336833000183, "train/lr": 0.00011922625236789447, "train/step_canary_loss": 11.9375, "perf/step_duration_sec": 4.607743468135595, "perf/samples_per_sec": 7.161857040915648, "perf/tokens_per_sec": 6096.259523615774, "perf/logical_batch_size": 33.0, "perf/logical_token_count": 28090.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.3670629406962955, "system/cuda_memory_allocated_gb": 16.247522354125977, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559020.410219, "event": "train_step", "step": 150, "epoch": 1, "metrics": {"train/step_loss": 2.48991414662954, "train/step_real_loss": 1.3496662974357605, "train/lr": 0.00011828373167982802, "train/step_canary_loss": 9.787500381469727, "perf/step_duration_sec": 5.096203599125147, "perf/samples_per_sec": 7.260306477227814, "perf/tokens_per_sec": 4987.634325356124, "perf/logical_batch_size": 37.0, "perf/logical_token_count": 25418.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.3912505604460925, "system/cuda_memory_allocated_gb": 16.6264328956604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559034.7611773, "event": "eval_step", "step": 150, "epoch": 1, "metrics": {"eval/loss": 0.8465344853360546, "eval/duration_sec": 14.343692186055705}}
{"timestamp": 1774559059.5274673, "event": "train_step", "step": 155, "epoch": 1, "metrics": {"train/step_loss": 1.6409546823212595, "train/step_real_loss": 1.4109845161437988, "train/lr": 0.00011730598739054154, "train/step_canary_loss": 9.0, "perf/step_duration_sec": 4.8965396150015295, "perf/samples_per_sec": 6.739453286336721, "perf/tokens_per_sec": 5674.415441238358, "perf/logical_batch_size": 33.0, "perf/logical_token_count": 27785.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.4154381801958893, "system/cuda_memory_allocated_gb": 16.247522354125977, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559086.7826245, "event": "train_step", "step": 160, "epoch": 1, "metrics": {"train/step_loss": 1.8622389400706572, "train/step_real_loss": 1.3360507488250732, "train/lr": 0.00011629370782916385, "train/step_canary_loss": 10.28125, "perf/step_duration_sec": 5.470216292887926, "perf/samples_per_sec": 6.215476350396771, "perf/tokens_per_sec": 5180.599538055709, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 28339.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.439625799945686, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559101.1275637, "event": "eval_step", "step": 160, "epoch": 1, "metrics": {"eval/loss": 0.8466571974144741, "eval/duration_sec": 14.328332623001188}}
{"timestamp": 1774559125.9494712, "event": "train_step", "step": 165, "epoch": 1, "metrics": {"train/step_loss": 1.7350735944860123, "train/step_real_loss": 1.241953194141388, "train/lr": 0.00011524760563755468, "train/step_canary_loss": 9.625, "perf/step_duration_sec": 5.42215427197516, "perf/samples_per_sec": 6.270570384861924, "perf/tokens_per_sec": 5020.144878704164, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 27220.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.4638134196954828, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559152.4945982, "event": "train_step", "step": 170, "epoch": 1, "metrics": {"train/step_loss": 2.093739182608468, "train/step_real_loss": 1.3447146713733673, "train/lr": 0.00011416841726860683, "train/step_canary_loss": 10.083333969116211, "perf/step_duration_sec": 4.8282034089788795, "perf/samples_per_sec": 7.249073213218699, "perf/tokens_per_sec": 5255.785195961077, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 25376.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.4874701420946528, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559166.7955632, "event": "eval_step", "step": 170, "epoch": 1, "metrics": {"eval/loss": 0.8467787429690361, "eval/duration_sec": 14.28548137517646}}
{"timestamp": 1774559191.9893062, "event": "train_step", "step": 175, "epoch": 1, "metrics": {"train/step_loss": 1.8259091096765854, "train/step_real_loss": 1.420497179031372, "train/lr": 0.00011305690246778548, "train/step_canary_loss": 8.3125, "perf/step_duration_sec": 5.374866001075134, "perf/samples_per_sec": 6.325739096230303, "perf/tokens_per_sec": 4558.997376883157, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 24504.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.509639334326591, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559218.1746178, "event": "train_step", "step": 180, "epoch": 1, "metrics": {"train/step_loss": 2.368437025282118, "train/step_real_loss": 1.3617572784423828, "train/lr": 0.00011191384373826953, "train/step_canary_loss": 10.421875, "perf/step_duration_sec": 4.852365015074611, "perf/samples_per_sec": 7.419062640209572, "perf/tokens_per_sec": 5502.265373082093, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 26699.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.5318085265585293, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559232.5191872, "event": "eval_step", "step": 180, "epoch": 1, "metrics": {"eval/loss": 0.8468846782025965, "eval/duration_sec": 14.328145914943889}}
{"timestamp": 1774559258.4477534, "event": "train_step", "step": 185, "epoch": 1, "metrics": {"train/step_loss": 2.2671251032087536, "train/step_real_loss": 1.263406366109848, "train/lr": 0.0001107400457900718, "train/step_canary_loss": 10.296875, "perf/step_duration_sec": 5.239895805018023, "perf/samples_per_sec": 6.870365621683613, "perf/tokens_per_sec": 5093.04020405196, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 26687.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.5539777187904675, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559283.2466376, "event": "train_step", "step": 190, "epoch": 1, "metrics": {"train/step_loss": 2.434699032757733, "train/step_real_loss": 1.268245667219162, "train/lr": 0.00010953633497352539, "train/step_canary_loss": 9.90000057220459, "perf/step_duration_sec": 4.863284363178536, "perf/samples_per_sec": 7.608027258315121, "perf/tokens_per_sec": 6162.502079235248, "perf/logical_batch_size": 37.0, "perf/logical_token_count": 29970.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.5761469110224058, "system/cuda_memory_allocated_gb": 16.6264328956604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559297.6711054, "event": "eval_step", "step": 190, "epoch": 1, "metrics": {"eval/loss": 0.847040081396699, "eval/duration_sec": 14.410996756982058}}
{"timestamp": 1774559323.4624872, "event": "train_step", "step": 195, "epoch": 1, "metrics": {"train/step_loss": 1.9696398053850446, "train/step_real_loss": 1.2089809775352478, "train/lr": 0.00010830355869753557, "train/step_canary_loss": 10.083333969116211, "perf/step_duration_sec": 5.458291776012629, "perf/samples_per_sec": 6.412262560571298, "perf/tokens_per_sec": 5479.919584264231, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 29911.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.5983161032543438, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559348.7154844, "event": "train_step", "step": 200, "epoch": 1, "metrics": {"train/step_loss": 1.5099033297914448, "train/step_real_loss": 1.2758378088474274, "train/lr": 0.00010704258483300614, "train/step_canary_loss": 9.0, "perf/step_duration_sec": 4.900695643154904, "perf/samples_per_sec": 6.7337379023104775, "perf/tokens_per_sec": 5855.903343045638, "perf/logical_batch_size": 33.0, "perf/logical_token_count": 28698.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.620485295486282, "system/cuda_memory_allocated_gb": 16.247522354125977, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559362.9754272, "event": "eval_step", "step": 200, "epoch": 1, "metrics": {"eval/loss": 0.8471845931966195, "eval/duration_sec": 14.248552802018821}}
{"timestamp": 1774559389.271019, "event": "train_step", "step": 205, "epoch": 1, "metrics": {"train/step_loss": 2.321085320578681, "train/step_real_loss": 1.2987209856510162, "train/lr": 0.0001057543011018612, "train/step_canary_loss": 10.5, "perf/step_duration_sec": 5.273128877859563, "perf/samples_per_sec": 6.827066213221951, "perf/tokens_per_sec": 4916.81515861682, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 25927.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.64265448771822, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559416.7390337, "event": "train_epoch", "step": 208, "epoch": 1, "metrics": {"train/epoch_loss": 2.128401287744091, "train/epoch_real_loss": 1.3418395462481736, "train/epoch_canary_loss": 9.875464486526338, "perf/epoch_duration_sec": 1368.1533101799432, "perf/epoch_samples_per_sec": 42.575637954155006, "perf/epoch_tokens_per_sec": 32259.456357412993, "perf/epoch_samples": 58250.0, "perf/epoch_tokens": 44135882.0, "system/cuda_epoch_peak_memory_gb": 74.68966627120972, "eval/loss": 0.8473063006319783, "eval/duration_sec": 14.333527690032497, "privacy/epsilon": 1.6559560030573832}}
{"timestamp": 1774559425.8804731, "event": "audit_epoch", "step": 208, "epoch": 1, "metrics": {"audit/delta": 1e-05, "audit/num_canaries": 500.0, "audit/num_members": 250.0, "audit/paper_guess_fraction": 0.2, "audit/paper_guess_steps": 20.0, "audit/loss/auc": 0.530384, "audit/loss/empirical_epsilon/0.05": 0.08366204984486103, "audit/loss/empirical_epsilon/0.01": 0.0, "audit/loss/empirical_epsilon_details/0.05/epsilon": 0.08366204984486103, "audit/loss/empirical_epsilon_details/0.05/num_guesses": 95.0, "audit/loss/empirical_epsilon_details/0.05/correct_guesses": 58.0, "audit/loss/empirical_epsilon_details/0.01/epsilon": 0.0, "audit/loss/empirical_epsilon_details/0.01/num_guesses": 0.0, "audit/loss/empirical_epsilon_details/0.01/correct_guesses": 0.0, "audit/embedding/auc": 0.531032, "audit/embedding/empirical_epsilon/0.05": 0.0, "audit/embedding/empirical_epsilon/0.01": 0.0, "audit/embedding/empirical_epsilon_details/0.05/epsilon": 0.0, "audit/embedding/empirical_epsilon_details/0.05/num_guesses": 0.0, "audit/embedding/empirical_epsilon_details/0.05/correct_guesses": 0.0, "audit/embedding/empirical_epsilon_details/0.01/epsilon": 0.0, "audit/embedding/empirical_epsilon_details/0.01/num_guesses": 0.0, "audit/embedding/empirical_epsilon_details/0.01/correct_guesses": 0.0, "perf/audit_duration_sec": 5.968299895059317}}
{"timestamp": 1774559436.0267518, "event": "train_step", "step": 210, "epoch": 2, "metrics": {"train/step_loss": 1.5706685384114583, "train/step_real_loss": 1.3189706802368164, "train/lr": 0.00010443961445209121, "train/step_canary_loss": 9.625, "perf/step_duration_sec": 4.8181796811986715, "perf/samples_per_sec": 6.84905964150142, "perf/tokens_per_sec": 5721.870462319777, "perf/logical_batch_size": 33.0, "perf/logical_token_count": 27569.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.664665370848081, "system/cuda_memory_allocated_gb": 16.247522354125977, "system/cuda_max_memory_allocated_gb": 74.68949747085571}}
{"timestamp": 1774559450.2979774, "event": "eval_step", "step": 210, "epoch": 2, "metrics": {"eval/loss": 0.8473422986201264, "eval/duration_sec": 14.265146326972172}}
{"timestamp": 1774559478.439248, "event": "train_step", "step": 215, "epoch": 2, "metrics": {"train/step_loss": 2.0742414474487303, "train/step_real_loss": 1.331201583147049, "train/lr": 0.0001030994504192644, "train/step_canary_loss": 10.0, "perf/step_duration_sec": 6.186825509183109, "perf/samples_per_sec": 5.657182338187731, "perf/tokens_per_sec": 4217.67188379059, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 26094.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.6846477710727013, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559503.3959737, "event": "train_step", "step": 220, "epoch": 2, "metrics": {"train/step_loss": 2.7938587540074398, "train/step_real_loss": 1.4212227761745453, "train/lr": 0.00010173475247495244, "train/step_canary_loss": 10.114583969116211, "perf/step_duration_sec": 4.850401181029156, "perf/samples_per_sec": 7.834403502255699, "perf/tokens_per_sec": 5575.827439960668, "perf/logical_batch_size": 38.0, "perf/logical_token_count": 27045.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.704630171297321, "system/cuda_memory_allocated_gb": 16.721160888671875, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559517.7483542, "event": "eval_step", "step": 220, "epoch": 2, "metrics": {"eval/loss": 0.8474802088669755, "eval/duration_sec": 14.33424908015877}}
{"timestamp": 1774559545.2226877, "event": "train_step", "step": 225, "epoch": 2, "metrics": {"train/step_loss": 1.996883283342634, "train/step_real_loss": 1.3247160613536835, "train/lr": 0.00010034648136252908, "train/step_canary_loss": 9.166666984558105, "perf/step_duration_sec": 5.59105938510038, "perf/samples_per_sec": 6.25999432116059, "perf/tokens_per_sec": 4943.070372969007, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 27637.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.724612571521941, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559570.7781942, "event": "train_step", "step": 230, "epoch": 2, "metrics": {"train/step_loss": 1.92525333516738, "train/step_real_loss": 1.4049566686153412, "train/lr": 9.893561442080963e-05, "train/step_canary_loss": 10.25, "perf/step_duration_sec": 5.317281369119883, "perf/samples_per_sec": 6.394245036092135, "perf/tokens_per_sec": 4311.789880661305, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 22927.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.7445949717465608, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559585.1297019, "event": "eval_step", "step": 230, "epoch": 2, "metrics": {"eval/loss": 0.8476190633055839, "eval/duration_sec": 14.340717867948115}}
{"timestamp": 1774559611.2625723, "event": "train_step", "step": 235, "epoch": 2, "metrics": {"train/step_loss": 2.3809074295891657, "train/step_real_loss": 1.3875052332878113, "train/lr": 9.750314489600751e-05, "train/step_canary_loss": 10.328125, "perf/step_duration_sec": 4.690604738891125, "perf/samples_per_sec": 7.674916562786428, "perf/tokens_per_sec": 5428.297931157445, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 25462.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.7645773719711806, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559636.4731991, "event": "train_step", "step": 240, "epoch": 2, "metrics": {"train/step_loss": 1.6647599538167317, "train/step_real_loss": 1.3749868273735046, "train/lr": 9.605008124249151e-05, "train/step_canary_loss": 10.9375, "perf/step_duration_sec": 4.822233859915286, "perf/samples_per_sec": 6.843301457092693, "perf/tokens_per_sec": 5682.221309872632, "perf/logical_batch_size": 33.0, "perf/logical_token_count": 27401.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.7845083173530645, "system/cuda_memory_allocated_gb": 16.247522354125977, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559650.8026834, "event": "eval_step", "step": 240, "epoch": 2, "metrics": {"eval/loss": 0.8477508156475695, "eval/duration_sec": 14.3184183628764}}
{"timestamp": 1774559676.9910817, "event": "train_step", "step": 245, "epoch": 2, "metrics": {"train/step_loss": 2.0103750228881836, "train/step_real_loss": 1.499304711818695, "train/lr": 9.457744641283705e-05, "train/step_canary_loss": 10.1875, "perf/step_duration_sec": 4.749070981983095, "perf/samples_per_sec": 7.159294971371945, "perf/tokens_per_sec": 5017.612937435943, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 23829.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.804262550512867, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559703.0872395, "event": "train_step", "step": 250, "epoch": 2, "metrics": {"train/step_loss": 2.474843437607224, "train/step_real_loss": 1.326381504535675, "train/lr": 9.308627713767046e-05, "train/step_canary_loss": 9.824999809265137, "perf/step_duration_sec": 5.456945129903033, "perf/samples_per_sec": 6.780350382716322, "perf/tokens_per_sec": 5356.476802345894, "perf/logical_batch_size": 37.0, "perf/logical_token_count": 29230.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.8238474343145514, "system/cuda_memory_allocated_gb": 16.6264328956604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559717.4105065, "event": "eval_step", "step": 250, "epoch": 2, "metrics": {"eval/loss": 0.8477847781032325, "eval/duration_sec": 14.306552066002041}}
{"timestamp": 1774559742.971318, "event": "train_step", "step": 255, "epoch": 2, "metrics": {"train/step_loss": 1.799149849835564, "train/step_real_loss": 1.3373779654502869, "train/lr": 9.157762319581367e-05, "train/step_canary_loss": 9.1875, "perf/step_duration_sec": 4.866647295886651, "perf/samples_per_sec": 6.986329177529921, "perf/tokens_per_sec": 5791.666888172304, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 28186.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.843239385504035, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559769.1780365, "event": "train_step", "step": 260, "epoch": 2, "metrics": {"train/step_loss": 2.103154046194894, "train/step_real_loss": 1.3960278034210205, "train/lr": 9.005254667524301e-05, "train/step_canary_loss": 9.645833969116211, "perf/step_duration_sec": 5.479270775103942, "perf/samples_per_sec": 6.387711328125786, "perf/tokens_per_sec": 5152.692969342152, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 28233.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.8624663041868499, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559783.5222437, "event": "eval_step", "step": 260, "epoch": 2, "metrics": {"eval/loss": 0.8478705518963662, "eval/duration_sec": 14.335434986045584}}
{"timestamp": 1774559807.9396706, "event": "train_step", "step": 265, "epoch": 2, "metrics": {"train/step_loss": 2.1713547842843193, "train/step_real_loss": 1.4491380453109741, "train/lr": 8.851212122538228e-05, "train/step_canary_loss": 9.875, "perf/step_duration_sec": 4.767590249190107, "perf/samples_per_sec": 7.341234915468169, "perf/tokens_per_sec": 5631.985677578165, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 26851.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.881539729973034, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559834.4034963, "event": "train_step", "step": 270, "epoch": 2, "metrics": {"train/step_loss": 2.3384444978502064, "train/step_real_loss": 1.386609435081482, "train/lr": 8.695743130125684e-05, "train/step_canary_loss": 9.953125, "perf/step_duration_sec": 5.385046212002635, "perf/samples_per_sec": 6.685179399159144, "perf/tokens_per_sec": 5081.479141071967, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 27364.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.900467909237016, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559848.772581, "event": "eval_step", "step": 270, "epoch": 2, "metrics": {"eval/loss": 0.8479423968290741, "eval/duration_sec": 14.35398749099113}}
{"timestamp": 1774559873.651874, "event": "train_step", "step": 275, "epoch": 2, "metrics": {"train/step_loss": 1.5575878836891868, "train/step_real_loss": 1.2937625050544739, "train/lr": 8.538957140004026e-05, "train/step_canary_loss": 10.0, "perf/step_duration_sec": 5.089407225139439, "perf/samples_per_sec": 6.484055714188968, "perf/tokens_per_sec": 5722.670384113749, "perf/logical_batch_size": 33.0, "perf/logical_token_count": 29125.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.9192594163827676, "system/cuda_memory_allocated_gb": 16.247522354125977, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559900.9281342, "event": "train_step", "step": 280, "epoch": 2, "metrics": {"train/step_loss": 2.2886579036712646, "train/step_real_loss": 1.3286463916301727, "train/lr": 8.380964529053162e-05, "train/step_canary_loss": 9.96875, "perf/step_duration_sec": 5.321737986057997, "perf/samples_per_sec": 6.7647073370229, "perf/tokens_per_sec": 4326.030342026144, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 23022.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.9378680010851501, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559915.2902293, "event": "eval_step", "step": 280, "epoch": 2, "metrics": {"eval/loss": 0.8480364416133274, "eval/duration_sec": 14.355422626016662}}
{"timestamp": 1774559939.9213169, "event": "train_step", "step": 285, "epoch": 2, "metrics": {"train/step_loss": 1.2202355933911873, "train/step_real_loss": 1.2583679556846619, "train/lr": 8.221876523610544e-05, "train/step_canary_loss": 0.0, "perf/step_duration_sec": 5.411638471996412, "perf/samples_per_sec": 5.913181408105936, "perf/tokens_per_sec": 4811.851370846206, "perf/logical_batch_size": 32.0, "perf/logical_token_count": 26040.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.956336606438703, "system/cuda_memory_allocated_gb": 16.909874439239502, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559966.647928, "event": "train_step", "step": 290, "epoch": 2, "metrics": {"train/step_loss": 2.077230943952288, "train/step_real_loss": 1.293455719947815, "train/lr": 8.061805121168158e-05, "train/step_canary_loss": 10.4375, "perf/step_duration_sec": 5.179025196004659, "perf/samples_per_sec": 6.758028523784867, "perf/tokens_per_sec": 5516.482140700962, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 28570.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.9746870226352882, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774559981.0026562, "event": "eval_step", "step": 290, "epoch": 2, "metrics": {"eval/loss": 0.8480958210473712, "eval/duration_sec": 14.343789409147575}}
{"timestamp": 1774560005.7170005, "event": "train_step", "step": 295, "epoch": 2, "metrics": {"train/step_loss": 1.872101334964528, "train/step_real_loss": 1.3445764183998108, "train/lr": 7.900863011526601e-05, "train/step_canary_loss": 10.3125, "perf/step_duration_sec": 5.2344231200404465, "perf/samples_per_sec": 6.495462674736405, "perf/tokens_per_sec": 5560.116049574363, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 29104.0, "perf/physical_batches": 5.0, "privacy/epsilon": 1.9929055810120353, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774560032.363273, "event": "train_step", "step": 300, "epoch": 2, "metrics": {"train/step_loss": 2.4508749472128377, "train/step_real_loss": 1.2615585327148438, "train/lr": 7.739163497461804e-05, "train/step_canary_loss": 10.0625, "perf/step_duration_sec": 4.758586144074798, "perf/samples_per_sec": 7.775418765103355, "perf/tokens_per_sec": 5754.650471988926, "perf/logical_batch_size": 37.0, "perf/logical_token_count": 27384.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.0109528801618994, "system/cuda_memory_allocated_gb": 16.6264328956604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774560046.677601, "event": "eval_step", "step": 300, "epoch": 2, "metrics": {"eval/loss": 0.8481822767379608, "eval/duration_sec": 14.289646297926083}}
{"timestamp": 1774560072.697593, "event": "train_step", "step": 305, "epoch": 2, "metrics": {"train/step_loss": 2.2884623209635415, "train/step_real_loss": 1.3635826110839844, "train/lr": 7.576820414960206e-05, "train/step_canary_loss": 9.6875, "perf/step_duration_sec": 5.614326270995662, "perf/samples_per_sec": 6.412167419977117, "perf/tokens_per_sec": 4795.410651334553, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 26923.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.0289020208489923, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774560098.999192, "event": "train_step", "step": 310, "epoch": 2, "metrics": {"train/step_loss": 1.7591877825119917, "train/step_real_loss": 1.328121393918991, "train/lr": 7.413948053078548e-05, "train/step_canary_loss": 8.65625, "perf/step_duration_sec": 4.873106176033616, "perf/samples_per_sec": 6.97706940333357, "perf/tokens_per_sec": 5514.552531640677, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 26873.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.046731649293609, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560113.278188, "event": "eval_step", "step": 310, "epoch": 2, "metrics": {"eval/loss": 0.8482703624123877, "eval/duration_sec": 14.186650173040107}}
{"timestamp": 1774560138.995842, "event": "train_step", "step": 315, "epoch": 2, "metrics": {"train/step_loss": 2.2547252443101673, "train/step_real_loss": 1.307073712348938, "train/lr": 7.250661073484708e-05, "train/step_canary_loss": 9.8359375, "perf/step_duration_sec": 4.812428422970697, "perf/samples_per_sec": 7.480630741054703, "perf/tokens_per_sec": 5734.319053615155, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 27596.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.0643990926316405, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560165.0429544, "event": "train_step", "step": 320, "epoch": 2, "metrics": {"train/step_loss": 2.66831300133153, "train/step_real_loss": 1.3561216294765472, "train/lr": 7.087074429736229e-05, "train/step_canary_loss": 9.666666984558105, "perf/step_duration_sec": 4.717704294016585, "perf/samples_per_sec": 8.054765121288971, "perf/tokens_per_sec": 5840.128647940888, "perf/logical_batch_size": 38.0, "perf/logical_token_count": 27552.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.081990142796045, "system/cuda_memory_allocated_gb": 16.721160888671875, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560179.3536744, "event": "eval_step", "step": 320, "epoch": 2, "metrics": {"eval/loss": 0.8482960782606493, "eval/duration_sec": 14.17633385094814}}
{"timestamp": 1774560204.8220775, "event": "train_step", "step": 325, "epoch": 2, "metrics": {"train/step_loss": 2.315086179309421, "train/step_real_loss": 1.3622844517230988, "train/lr": 6.923303286353341e-05, "train/step_canary_loss": 9.9375, "perf/step_duration_sec": 5.28622017102316, "perf/samples_per_sec": 6.810159023897054, "perf/tokens_per_sec": 5095.133976378979, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 26934.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.0994319316048977, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560230.3968325, "event": "train_step", "step": 330, "epoch": 2, "metrics": {"train/step_loss": 2.083483968462263, "train/step_real_loss": 1.4838886857032776, "train/lr": 6.759462937743482e-05, "train/step_canary_loss": 8.479166984558105, "perf/step_duration_sec": 4.692184546031058, "perf/samples_per_sec": 7.45921215515813, "perf/tokens_per_sec": 6244.639295951097, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 29301.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.1167645390743552, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560244.6999924, "event": "eval_step", "step": 330, "epoch": 2, "metrics": {"eval/loss": 0.8483132300051777, "eval/duration_sec": 14.26540448796004}}
{"timestamp": 1774560270.3830054, "event": "train_step", "step": 335, "epoch": 2, "metrics": {"train/step_loss": 1.8987440501942354, "train/step_real_loss": 1.3846030533313751, "train/lr": 6.595668727034364e-05, "train/step_canary_loss": 10.125, "perf/step_duration_sec": 5.418661136878654, "perf/samples_per_sec": 6.27461270249965, "perf/tokens_per_sec": 5138.169613617508, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 27842.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.1340128150355056, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560295.6445541, "event": "train_step", "step": 340, "epoch": 2, "metrics": {"train/step_loss": 1.8939583722282858, "train/step_real_loss": 1.3150651454925537, "train/lr": 6.432035964872754e-05, "train/step_canary_loss": 11.15625, "perf/step_duration_sec": 4.84555484703742, "perf/samples_per_sec": 7.016740305970875, "perf/tokens_per_sec": 5136.6666534004435, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 24890.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.1510995008588942, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560310.057485, "event": "eval_step", "step": 340, "epoch": 2, "metrics": {"eval/loss": 0.8483494562520222, "eval/duration_sec": 14.328203893033788}}
{"timestamp": 1774560337.3352282, "event": "train_step", "step": 345, "epoch": 2, "metrics": {"train/step_loss": 1.5868641535441081, "train/step_real_loss": 1.2692661583423615, "train/lr": 6.268679848246117e-05, "train/step_canary_loss": 11.75, "perf/step_duration_sec": 5.412797935074195, "perf/samples_per_sec": 6.096662095247355, "perf/tokens_per_sec": 4737.660690015854, "perf/logical_batch_size": 33.0, "perf/logical_token_count": 25644.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.1681463057336154, "system/cuda_memory_allocated_gb": 16.247522354125977, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560362.704056, "event": "train_step", "step": 350, "epoch": 2, "metrics": {"train/step_loss": 2.4502257012032174, "train/step_real_loss": 1.266667127609253, "train/lr": 6.105715379384262e-05, "train/step_canary_loss": 10.02500057220459, "perf/step_duration_sec": 4.753133960068226, "perf/samples_per_sec": 7.784337725560108, "perf/tokens_per_sec": 5776.8201423899845, "perf/logical_batch_size": 37.0, "perf/logical_token_count": 27458.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.1850083661862754, "system/cuda_memory_allocated_gb": 16.6264328956604, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560377.1350248, "event": "eval_step", "step": 350, "epoch": 2, "metrics": {"eval/loss": 0.848394123498689, "eval/duration_sec": 14.286400699988008}}
{"timestamp": 1774560402.8301609, "event": "train_step", "step": 355, "epoch": 2, "metrics": {"train/step_loss": 2.3293143113454184, "train/step_real_loss": 1.4466504752635956, "train/lr": 5.9432572847981164e-05, "train/step_canary_loss": 9.390625, "perf/step_duration_sec": 4.793874611146748, "perf/samples_per_sec": 7.509583149357425, "perf/tokens_per_sec": 5711.246584618249, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 27379.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.2018356969074238, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560429.3453176, "event": "train_step", "step": 360, "epoch": 2, "metrics": {"train/step_loss": 2.275719854566786, "train/step_real_loss": 1.3941692113876343, "train/lr": 5.781419934512588e-05, "train/step_canary_loss": 9.328125, "perf/step_duration_sec": 5.395623786840588, "perf/samples_per_sec": 6.672073780940875, "perf/tokens_per_sec": 4761.265984232529, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 25690.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.218505053207597, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560443.6799695, "event": "eval_step", "step": 360, "epoch": 2, "metrics": {"eval/loss": 0.8483872571113433, "eval/duration_sec": 14.229980896925554}}
{"timestamp": 1774560469.781411, "event": "train_step", "step": 365, "epoch": 2, "metrics": {"train/step_loss": 1.510118917985396, "train/step_real_loss": 1.2448101341724396, "train/lr": 5.6203172615504136e-05, "train/step_canary_loss": 10.0, "perf/step_duration_sec": 4.7617223400156945, "perf/samples_per_sec": 6.930265488745661, "perf/tokens_per_sec": 5727.549414380619, "perf/logical_batch_size": 33.0, "perf/logical_token_count": 27273.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.2351271767123455, "system/cuda_memory_allocated_gb": 16.247522354125977, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560496.934338, "event": "train_step", "step": 370, "epoch": 2, "metrics": {"train/step_loss": 2.0685702732631137, "train/step_real_loss": 1.3152330815792084, "train/lr": 5.460062681723634e-05, "train/step_canary_loss": 10.104166984558105, "perf/step_duration_sec": 5.522874823072925, "perf/samples_per_sec": 6.337279246992966, "perf/tokens_per_sec": 4886.766559916948, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 26989.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.251604096938732, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560511.2665586, "event": "eval_step", "step": 370, "epoch": 2, "metrics": {"eval/loss": 0.8484010513533247, "eval/duration_sec": 14.218599178129807}}
{"timestamp": 1774560537.5788438, "event": "train_step", "step": 375, "epoch": 2, "metrics": {"train/step_loss": 1.8943779328290153, "train/step_real_loss": 1.3877765536308289, "train/lr": 5.300769013789195e-05, "train/step_canary_loss": 10.0, "perf/step_duration_sec": 5.705397455953062, "perf/samples_per_sec": 5.959269316903436, "perf/tokens_per_sec": 4729.731838724682, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 26985.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.268035937304379, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560563.3753562, "event": "train_step", "step": 380, "epoch": 2, "metrics": {"train/step_loss": 2.207952690124512, "train/step_real_loss": 1.43447944521904, "train/lr": 5.142548400024875e-05, "train/step_canary_loss": 10.458333969116211, "perf/step_duration_sec": 5.331795918988064, "perf/samples_per_sec": 6.564392285787777, "perf/tokens_per_sec": 5117.037563804227, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 27283.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.284320688797799, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560577.6584978, "event": "eval_step", "step": 380, "epoch": 2, "metrics": {"eval/loss": 0.848428938029842, "eval/duration_sec": 14.268167359055951}}
{"timestamp": 1774560603.255523, "event": "train_step", "step": 385, "epoch": 2, "metrics": {"train/step_loss": 1.5885022192290335, "train/step_real_loss": 1.3217366635799408, "train/lr": 4.985512227281438e-05, "train/step_canary_loss": 10.125, "perf/step_duration_sec": 4.661598626989871, "perf/samples_per_sec": 7.079116552192108, "perf/tokens_per_sec": 5287.241989753906, "perf/logical_batch_size": 33.0, "perf/logical_token_count": 24647.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.3005778698106965, "system/cuda_memory_allocated_gb": 16.247522354125977, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560629.0995705, "event": "train_step", "step": 390, "epoch": 2, "metrics": {"train/step_loss": 1.7334133316488827, "train/step_real_loss": 1.2655797898769379, "train/lr": 4.829771048566633e-05, "train/step_canary_loss": 9.21875, "perf/step_duration_sec": 4.860105969943106, "perf/samples_per_sec": 6.995732235113798, "perf/tokens_per_sec": 5779.709367186664, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 28090.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.316670719177237, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560643.4201152, "event": "eval_step", "step": 390, "epoch": 2, "metrics": {"eval/loss": 0.84846505404196, "eval/duration_sec": 14.308998527005315}}
{"timestamp": 1774560668.4482021, "event": "train_step", "step": 395, "epoch": 2, "metrics": {"train/step_loss": 1.8592574736651253, "train/step_real_loss": 1.3328829407691956, "train/lr": 4.6754345052161783e-05, "train/step_canary_loss": 10.28125, "perf/step_duration_sec": 4.722605478018522, "perf/samples_per_sec": 7.199415695055155, "perf/tokens_per_sec": 5434.076617123542, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 25663.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.332763568543778, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560694.6040494, "event": "train_step", "step": 400, "epoch": 2, "metrics": {"train/step_loss": 2.2962468121502853, "train/step_real_loss": 1.2741759121418, "train/lr": 4.5226112497066e-05, "train/step_canary_loss": 8.83750057220459, "perf/step_duration_sec": 5.1107431498821825, "perf/samples_per_sec": 7.2396516347828905, "perf/tokens_per_sec": 5299.620662921475, "perf/logical_batch_size": 37.0, "perf/logical_token_count": 27085.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.348670823328917, "system/cuda_memory_allocated_gb": 16.6264328956604, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560708.9928586, "event": "eval_step", "step": 400, "epoch": 2, "metrics": {"eval/loss": 0.8484586563638665, "eval/duration_sec": 14.375536187086254}}
{"timestamp": 1774560734.8393927, "event": "train_step", "step": 405, "epoch": 2, "metrics": {"train/step_loss": 1.5759954741506865, "train/step_real_loss": 1.3381359577178955, "train/lr": 4.3714088691641896e-05, "train/step_canary_loss": 9.1875, "perf/step_duration_sec": 5.455053962999955, "perf/samples_per_sec": 6.049436031949345, "perf/tokens_per_sec": 5235.145278800285, "perf/logical_batch_size": 33.0, "perf/logical_token_count": 28558.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.3645720364430685, "system/cuda_memory_allocated_gb": 16.247522354125977, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560761.2520165, "event": "train_step", "step": 410, "epoch": 2, "metrics": {"train/step_loss": 2.752719853375409, "train/step_real_loss": 1.6164259910583496, "train/lr": 4.221933809623992e-05, "train/step_canary_loss": 10.02500057220459, "perf/step_duration_sec": 5.600042568985373, "perf/samples_per_sec": 6.607092632637565, "perf/tokens_per_sec": 4382.645256292534, "perf/logical_batch_size": 37.0, "perf/logical_token_count": 24543.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.3803384308636324, "system/cuda_memory_allocated_gb": 16.6264328956604, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560775.643603, "event": "eval_step", "step": 410, "epoch": 2, "metrics": {"eval/loss": 0.8484945239668544, "eval/duration_sec": 14.379757352173328}}
{"timestamp": 1774560801.2065222, "event": "train_step", "step": 415, "epoch": 2, "metrics": {"train/step_loss": 1.9820210865565708, "train/step_real_loss": 1.3397105038166046, "train/lr": 4.074291301092102e-05, "train/step_canary_loss": 8.833333969116211, "perf/step_duration_sec": 5.44986342987977, "perf/samples_per_sec": 6.422179280329625, "perf/tokens_per_sec": 4621.400210125198, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 25186.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.396048272871404, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68969058990479}}
{"timestamp": 1774560817.8175077, "event": "train_epoch", "step": 416, "epoch": 2, "metrics": {"train/epoch_loss": 2.113187212149485, "train/epoch_real_loss": 1.344985583841872, "train/epoch_canary_loss": 9.706668990476329, "perf/epoch_duration_sec": 1377.3823250350542, "perf/epoch_samples_per_sec": 42.224296727867376, "perf/epoch_tokens_per_sec": 32039.881881653204, "perf/epoch_samples": 58159.0, "perf/epoch_tokens": 44131167.0, "system/cuda_epoch_peak_memory_gb": 74.68969058990479, "eval/loss": 0.8484931107271801, "eval/duration_sec": 14.344455992104486, "privacy/epsilon": 2.3991902412729584}}
{"timestamp": 1774560826.9155543, "event": "audit_epoch", "step": 416, "epoch": 2, "metrics": {"audit/delta": 1e-05, "audit/num_canaries": 500.0, "audit/num_members": 250.0, "audit/paper_guess_fraction": 0.2, "audit/paper_guess_steps": 20.0, "audit/loss/auc": 0.529296, "audit/loss/empirical_epsilon/0.05": 0.09125839080661535, "audit/loss/empirical_epsilon/0.01": 0.0, "audit/loss/empirical_epsilon_details/0.05/epsilon": 0.09125839080661535, "audit/loss/empirical_epsilon_details/0.05/num_guesses": 100.0, "audit/loss/empirical_epsilon_details/0.05/correct_guesses": 61.0, "audit/loss/empirical_epsilon_details/0.01/epsilon": 0.0, "audit/loss/empirical_epsilon_details/0.01/num_guesses": 0.0, "audit/loss/empirical_epsilon_details/0.01/correct_guesses": 0.0, "audit/embedding/auc": 0.542392, "audit/embedding/empirical_epsilon/0.05": 0.0, "audit/embedding/empirical_epsilon/0.01": 0.0, "audit/embedding/empirical_epsilon_details/0.05/epsilon": 0.0, "audit/embedding/empirical_epsilon_details/0.05/num_guesses": 0.0, "audit/embedding/empirical_epsilon_details/0.05/correct_guesses": 0.0, "audit/embedding/empirical_epsilon_details/0.01/epsilon": 0.0, "audit/embedding/empirical_epsilon_details/0.01/num_guesses": 0.0, "audit/embedding/empirical_epsilon_details/0.01/correct_guesses": 0.0, "perf/audit_duration_sec": 5.908021365990862}}
{"timestamp": 1774560850.0085697, "event": "train_step", "step": 420, "epoch": 3, "metrics": {"train/step_loss": 1.1538096052227598, "train/step_real_loss": 1.189866155385971, "train/lr": 3.928585283464046e-05, "train/step_canary_loss": 0.0, "perf/step_duration_sec": 5.7631713869050145, "perf/samples_per_sec": 5.55249841653328, "perf/tokens_per_sec": 4910.317271546102, "perf/logical_batch_size": 32.0, "perf/logical_token_count": 28299.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.411691819202272, "system/cuda_memory_allocated_gb": 16.909874439239502, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774560864.2485898, "event": "eval_step", "step": 420, "epoch": 3, "metrics": {"eval/loss": 0.8485084484246643, "eval/duration_sec": 14.22628792701289}}
{"timestamp": 1774560890.540943, "event": "train_step", "step": 425, "epoch": 3, "metrics": {"train/step_loss": 1.7848442863015568, "train/step_real_loss": 1.269443929195404, "train/lr": 3.7849183333514016e-05, "train/step_canary_loss": 10.03125, "perf/step_duration_sec": 4.773140619974583, "perf/samples_per_sec": 7.123192611949707, "perf/tokens_per_sec": 6033.134636656345, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 28797.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.427210554524291, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774560916.9385135, "event": "train_step", "step": 430, "epoch": 3, "metrics": {"train/step_loss": 1.8522330171921675, "train/step_real_loss": 1.3215132057666779, "train/lr": 3.6433915918681605e-05, "train/step_canary_loss": 10.34375, "perf/step_duration_sec": 4.7064781279768795, "perf/samples_per_sec": 7.224085414928974, "perf/tokens_per_sec": 5170.7453722856335, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 24336.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.44272928984631, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774560931.3039591, "event": "eval_step", "step": 430, "epoch": 3, "metrics": {"eval/loss": 0.8485086685554548, "eval/duration_sec": 14.341718771960586}}
{"timestamp": 1774560956.9222178, "event": "train_step", "step": 435, "epoch": 3, "metrics": {"train/step_loss": 2.1504132407052174, "train/step_real_loss": 1.526819109916687, "train/lr": 3.5041046934276864e-05, "train/step_canary_loss": 8.802083969116211, "perf/step_duration_sec": 5.123670374974608, "perf/samples_per_sec": 6.831040531207758, "perf/tokens_per_sec": 4801.245630506024, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 24600.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.458078063684959, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774560983.1538332, "event": "train_step", "step": 440, "epoch": 3, "metrics": {"train/step_loss": 1.7019892019384049, "train/step_real_loss": 1.319105714559555, "train/lr": 3.367155695600353e-05, "train/step_canary_loss": 7.828125, "perf/step_duration_sec": 5.176391114946455, "perf/samples_per_sec": 6.568282659675282, "perf/tokens_per_sec": 5070.327843701102, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 26246.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.4734059560195463, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774560997.4815164, "event": "eval_step", "step": 440, "epoch": 3, "metrics": {"eval/loss": 0.848517840389501, "eval/duration_sec": 14.306687124073505}}
{"timestamp": 1774561022.892964, "event": "train_step", "step": 445, "epoch": 3, "metrics": {"train/step_loss": 2.50487706467912, "train/step_real_loss": 1.336693823337555, "train/lr": 3.232641010081339e-05, "train/step_canary_loss": 9.981249809265137, "perf/step_duration_sec": 5.4494336121715605, "perf/samples_per_sec": 6.789696440628031, "perf/tokens_per_sec": 4964.552635263534, "perf/logical_batch_size": 37.0, "perf/logical_token_count": 27054.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.4886709507369824, "system/cuda_memory_allocated_gb": 16.6264328956604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561049.7502263, "event": "train_step", "step": 450, "epoch": 3, "metrics": {"train/step_loss": 1.8840222078211166, "train/step_real_loss": 1.3533360958099365, "train/lr": 3.100655334817044e-05, "train/step_canary_loss": 10.375, "perf/step_duration_sec": 5.34880125708878, "perf/samples_per_sec": 6.356564464783527, "perf/tokens_per_sec": 4939.6114624719285, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 26421.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.5038082630632235, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561064.0000763, "event": "eval_step", "step": 450, "epoch": 3, "metrics": {"eval/loss": 0.8485311972485348, "eval/duration_sec": 14.237569699063897}}
{"timestamp": 1774561089.179267, "event": "train_step", "step": 455, "epoch": 3, "metrics": {"train/step_loss": 2.8950751622517905, "train/step_real_loss": 1.3213415741920471, "train/lr": 2.9712915873380282e-05, "train/step_canary_loss": 10.089285850524902, "perf/step_duration_sec": 4.954125403193757, "perf/samples_per_sec": 7.872227048362163, "perf/tokens_per_sec": 5379.355149714144, "perf/logical_batch_size": 39.0, "perf/logical_token_count": 26650.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.5189455753894645, "system/cuda_memory_allocated_gb": 16.81588888168335, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561116.025384, "event": "train_step", "step": 460, "epoch": 3, "metrics": {"train/step_loss": 2.5384145685144373, "train/step_real_loss": 1.4233230650424957, "train/lr": 2.8446408393453304e-05, "train/step_canary_loss": 9.675000190734863, "perf/step_duration_sec": 5.59700956591405, "perf/samples_per_sec": 6.610672996760819, "perf/tokens_per_sec": 4608.711079768792, "perf/logical_batch_size": 37.0, "perf/logical_token_count": 25795.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.533957396449435, "system/cuda_memory_allocated_gb": 16.6264328956604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561130.337623, "event": "eval_step", "step": 460, "epoch": 3, "metrics": {"eval/loss": 0.848521682687781, "eval/duration_sec": 14.294991586124524}}
{"timestamp": 1774561155.2550473, "event": "train_step", "step": 465, "epoch": 3, "metrics": {"train/step_loss": 1.7144349322599524, "train/step_real_loss": 1.2434621155261993, "train/lr": 2.7207922525962636e-05, "train/step_canary_loss": 9.25, "perf/step_duration_sec": 5.018907584017143, "perf/samples_per_sec": 6.774382558522095, "perf/tokens_per_sec": 5839.916258537724, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 29310.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.548904391030205, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561181.5087345, "event": "train_step", "step": 470, "epoch": 3, "metrics": {"train/step_loss": 2.050832666669573, "train/step_real_loss": 1.3856762945652008, "train/lr": 2.5998330161347976e-05, "train/step_canary_loss": 9.145833969116211, "perf/step_duration_sec": 5.241146533982828, "perf/samples_per_sec": 6.677928154281723, "perf/tokens_per_sec": 5364.475085308141, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 28116.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.5638513856109753, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561195.7935584, "event": "eval_step", "step": 470, "epoch": 3, "metrics": {"eval/loss": 0.8485205579887737, "eval/duration_sec": 14.271867047064006}}
{"timestamp": 1774561221.6017852, "event": "train_step", "step": 475, "epoch": 3, "metrics": {"train/step_loss": 2.624330391754975, "train/step_real_loss": 1.2882882356643677, "train/lr": 2.4818482849107253e-05, "train/step_canary_loss": 11.175000190734863, "perf/step_duration_sec": 5.4966746747959405, "perf/samples_per_sec": 6.731342527811798, "perf/tokens_per_sec": 5186.408453590777, "perf/logical_batch_size": 37.0, "perf/logical_token_count": 28508.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.5786325892580035, "system/cuda_memory_allocated_gb": 16.6264328956604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561246.803026, "event": "train_step", "step": 480, "epoch": 3, "metrics": {"train/step_loss": 2.0559374491373696, "train/step_real_loss": 1.234804630279541, "train/lr": 2.366921119830816e-05, "train/step_canary_loss": 8.625, "perf/step_duration_sec": 4.680067221866921, "perf/samples_per_sec": 7.692197204304104, "perf/tokens_per_sec": 6025.554476704881, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 28200.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.59338952764301, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561261.163607, "event": "eval_step", "step": 480, "epoch": 3, "metrics": {"eval/loss": 0.8485760211267255, "eval/duration_sec": 14.347748975036666}}
{"timestamp": 1774561288.4332306, "event": "train_step", "step": 485, "epoch": 3, "metrics": {"train/step_loss": 1.8063245661118452, "train/step_real_loss": 1.2981261014938354, "train/lr": 2.2551324292841998e-05, "train/step_canary_loss": 9.9375, "perf/step_duration_sec": 5.603643656941131, "perf/samples_per_sec": 6.06748074672536, "perf/tokens_per_sec": 4733.170348394021, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 26523.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.608146466028017, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561314.239187, "event": "train_step", "step": 490, "epoch": 3, "metrics": {"train/step_loss": 2.012660162789481, "train/step_real_loss": 1.3185345232486725, "train/lr": 2.1465609121830783e-05, "train/step_canary_loss": 9.416666984558105, "perf/step_duration_sec": 5.054407468996942, "perf/samples_per_sec": 6.924649469732171, "perf/tokens_per_sec": 5533.784161954538, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 27970.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.622720795991837, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561328.5810933, "event": "eval_step", "step": 490, "epoch": 3, "metrics": {"eval/loss": 0.8485268791290846, "eval/duration_sec": 14.324147312901914}}
{"timestamp": 1774561354.565853, "event": "train_step", "step": 495, "epoch": 3, "metrics": {"train/step_loss": 2.2673527664608426, "train/step_real_loss": 1.3222562372684479, "train/lr": 2.04128300255893e-05, "train/step_canary_loss": 9.828125, "perf/step_duration_sec": 5.36041372991167, "perf/samples_per_sec": 6.715899520799343, "perf/tokens_per_sec": 5078.525907075569, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 27223.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.637287939020626, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561380.6077175, "event": "train_step", "step": 500, "epoch": 3, "metrics": {"train/step_loss": 2.4888906736631653, "train/step_real_loss": 1.430514246225357, "train/lr": 1.939372815753177e-05, "train/step_canary_loss": 9.262499809265137, "perf/step_duration_sec": 5.711737324018031, "perf/samples_per_sec": 6.47788893309464, "perf/tokens_per_sec": 4470.968910390212, "perf/logical_batch_size": 37.0, "perf/logical_token_count": 25537.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.6518550820494142, "system/cuda_memory_allocated_gb": 16.6264328956604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561394.8906922, "event": "eval_step", "step": 500, "epoch": 3, "metrics": {"eval/loss": 0.8485470659692178, "eval/duration_sec": 14.269119519973174}}
{"timestamp": 1774561421.3410587, "event": "train_step", "step": 505, "epoch": 3, "metrics": {"train/step_loss": 1.966035983141731, "train/step_real_loss": 1.366256982088089, "train/lr": 1.840902096240187e-05, "train/step_canary_loss": 11.5625, "perf/step_duration_sec": 5.488221630919725, "perf/samples_per_sec": 6.195085090669385, "perf/tokens_per_sec": 4572.883838839104, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 25097.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.666247556144702, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561446.9770482, "event": "train_step", "step": 510, "epoch": 3, "metrics": {"train/step_loss": 1.8509371701408834, "train/step_real_loss": 1.4021676182746887, "train/lr": 1.745940167119394e-05, "train/step_canary_loss": 9.03125, "perf/step_duration_sec": 4.957968695089221, "perf/samples_per_sec": 6.857647171850115, "perf/tokens_per_sec": 5265.664550455616, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 26107.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.680625163949644, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561461.3933687, "event": "eval_step", "step": 510, "epoch": 3, "metrics": {"eval/loss": 0.8485351553694769, "eval/duration_sec": 14.385921356966719}}
{"timestamp": 1774561487.2952607, "event": "train_step", "step": 515, "epoch": 3, "metrics": {"train/step_loss": 1.8762224982766544, "train/step_real_loss": 1.3645801544189453, "train/lr": 1.654553881312016e-05, "train/step_canary_loss": 10.0625, "perf/step_duration_sec": 4.9261083118617535, "perf/samples_per_sec": 6.902000087600627, "perf/tokens_per_sec": 4559.177057865367, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 22459.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.6950027717545866, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561513.949884, "event": "train_step", "step": 520, "epoch": 3, "metrics": {"train/step_loss": 2.5595649513038428, "train/step_real_loss": 1.3852782547473907, "train/lr": 1.5668075744968074e-05, "train/step_canary_loss": 10.074999809265137, "perf/step_duration_sec": 5.4532803618349135, "perf/samples_per_sec": 6.784906981666771, "perf/tokens_per_sec": 4867.162192091877, "perf/logical_batch_size": 37.0, "perf/logical_token_count": 26542.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.7092397760930864, "system/cuda_memory_allocated_gb": 16.6264328956604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561528.3120036, "event": "eval_step", "step": 520, "epoch": 3, "metrics": {"eval/loss": 0.8485690009864895, "eval/duration_sec": 14.35072849295102}}
{"timestamp": 1774561554.5826762, "event": "train_step", "step": 525, "epoch": 3, "metrics": {"train/step_loss": 2.0029184886387417, "train/step_real_loss": 1.2727233171463013, "train/lr": 1.4827630198179268e-05, "train/step_canary_loss": 9.791666984558105, "perf/step_duration_sec": 5.13253407808952, "perf/samples_per_sec": 6.819243568087136, "perf/tokens_per_sec": 5268.742416234638, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 27042.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.723428108102348, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561580.2294695, "event": "train_step", "step": 530, "epoch": 3, "metrics": {"train/step_loss": 1.6140749382250237, "train/step_real_loss": 1.3832647800445557, "train/lr": 1.4024793843968244e-05, "train/step_canary_loss": 9.0, "perf/step_duration_sec": 5.246445879107341, "perf/samples_per_sec": 6.289972442375561, "perf/tokens_per_sec": 4624.08277127367, "perf/logical_batch_size": 33.0, "perf/logical_token_count": 24260.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.7376164401116103, "system/cuda_memory_allocated_gb": 16.247522354125977, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561594.527408, "event": "eval_step", "step": 530, "epoch": 3, "metrics": {"eval/loss": 0.8485432566905564, "eval/duration_sec": 14.285051121143624}}
{"timestamp": 1774561621.0193505, "event": "train_step", "step": 535, "epoch": 3, "metrics": {"train/step_loss": 2.6340620643214177, "train/step_real_loss": 1.3134954571723938, "train/lr": 1.326013187678748e-05, "train/step_canary_loss": 9.677083969116211, "perf/step_duration_sec": 4.868667971109971, "perf/samples_per_sec": 7.805009547885982, "perf/tokens_per_sec": 5207.584528344768, "perf/logical_batch_size": 38.0, "perf/logical_token_count": 25354.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.751725831720154, "system/cuda_memory_allocated_gb": 16.721160888671875, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561646.8280683, "event": "train_step", "step": 540, "epoch": 3, "metrics": {"train/step_loss": 2.186668450491769, "train/step_real_loss": 1.4072936177253723, "train/lr": 1.2534182616432408e-05, "train/step_canary_loss": 10.5, "perf/step_duration_sec": 5.280470572877675, "perf/samples_per_sec": 6.628197149657857, "perf/tokens_per_sec": 4709.618140424036, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 24869.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.765725146660663, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561661.1712415, "event": "eval_step", "step": 540, "epoch": 3, "metrics": {"eval/loss": 0.8485460245812482, "eval/duration_sec": 14.332059542881325}}
{"timestamp": 1774561687.0665293, "event": "train_step", "step": 545, "epoch": 3, "metrics": {"train/step_loss": 2.270850578943888, "train/step_real_loss": 1.3535350263118744, "train/lr": 1.184745712906565e-05, "train/step_canary_loss": 9.609375, "perf/step_duration_sec": 5.501790995942429, "perf/samples_per_sec": 6.543323806111501, "perf/tokens_per_sec": 4993.283100041533, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 27472.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.779724461601172, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561712.6492045, "event": "train_step", "step": 550, "epoch": 3, "metrics": {"train/step_loss": 2.1144141605922155, "train/step_real_loss": 1.3595154583454132, "train/lr": 1.1200438867428068e-05, "train/step_canary_loss": 10.166666984558105, "perf/step_duration_sec": 5.274112347979099, "perf/samples_per_sec": 6.636187796304924, "perf/tokens_per_sec": 4295.88876859659, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 22657.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.7937237765416802, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561726.9447677, "event": "eval_step", "step": 550, "epoch": 3, "metrics": {"eval/loss": 0.8485449015755545, "eval/duration_sec": 14.283992771990597}}
{"timestamp": 1774561752.2924328, "event": "train_step", "step": 555, "epoch": 3, "metrics": {"train/step_loss": 1.8974150208865894, "train/step_real_loss": 1.3773315846920013, "train/lr": 1.0593583330489345e-05, "train/step_canary_loss": 10.21875, "perf/step_duration_sec": 4.863103995099664, "perf/samples_per_sec": 6.991419479052948, "perf/tokens_per_sec": 5288.186315964843, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 25717.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.807546236227026, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561778.4963992, "event": "train_step", "step": 560, "epoch": 3, "metrics": {"train/step_loss": 1.6954598571314956, "train/step_real_loss": 1.4671929776668549, "train/lr": 1.002731774277807e-05, "train/step_canary_loss": 9.0, "perf/step_duration_sec": 5.632957021007314, "perf/samples_per_sec": 5.858379511317267, "perf/tokens_per_sec": 4374.611755087276, "perf/logical_batch_size": 33.0, "perf/logical_token_count": 24642.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.821356792127403, "system/cuda_memory_allocated_gb": 16.247522354125977, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561792.7788231, "event": "eval_step", "step": 560, "epoch": 3, "metrics": {"eval/loss": 0.8485205733979291, "eval/duration_sec": 14.27125578187406}}
{"timestamp": 1774561817.9552119, "event": "train_step", "step": 565, "epoch": 3, "metrics": {"train/step_loss": 2.1035734176635743, "train/step_real_loss": 1.318361520767212, "train/lr": 9.502040753616962e-06, "train/step_canary_loss": 10.479166984558105, "perf/step_duration_sec": 4.720770954852924, "perf/samples_per_sec": 7.414043243089397, "perf/tokens_per_sec": 5555.448516950358, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 26226.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.8351673480277806, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561844.4099154, "event": "train_step", "step": 570, "epoch": 3, "metrics": {"train/step_loss": 2.738384849146793, "train/step_real_loss": 1.3201912641525269, "train/lr": 9.018122156474845e-06, "train/step_canary_loss": 10.302083969116211, "perf/step_duration_sec": 5.379596892977133, "perf/samples_per_sec": 7.063726289530654, "perf/tokens_per_sec": 5538.704961127773, "perf/logical_batch_size": 38.0, "perf/logical_token_count": 29796.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.8489230370212, "system/cuda_memory_allocated_gb": 16.721160888671875, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561858.7876792, "event": "eval_step", "step": 570, "epoch": 3, "metrics": {"eval/loss": 0.8485367150807923, "eval/duration_sec": 14.366766993887722}}
{"timestamp": 1774561885.2539098, "event": "train_step", "step": 575, "epoch": 3, "metrics": {"train/step_loss": 2.268237484825982, "train/step_real_loss": 1.2841890454292297, "train/lr": 8.575902628633133e-06, "train/step_canary_loss": 10.140625, "perf/step_duration_sec": 5.593655822100118, "perf/samples_per_sec": 6.435862545880402, "perf/tokens_per_sec": 4661.709770732705, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 26076.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.862545091214684, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561912.0462737, "event": "train_step", "step": 580, "epoch": 3, "metrics": {"train/step_loss": 1.9337453297206333, "train/step_real_loss": 1.216596394777298, "train/lr": 8.175693491350082e-06, "train/step_canary_loss": 9.583333969116211, "perf/step_duration_sec": 5.297230056952685, "perf/samples_per_sec": 6.607226724854442, "perf/tokens_per_sec": 5094.360582769199, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 26986.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.8761671454081683, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561926.4586546, "event": "eval_step", "step": 580, "epoch": 3, "metrics": {"eval/loss": 0.8485535840419206, "eval/duration_sec": 14.398916372098029}}
{"timestamp": 1774561953.0506368, "event": "train_step", "step": 585, "epoch": 3, "metrics": {"train/step_loss": 1.8880423938526827, "train/step_real_loss": 1.3126856684684753, "train/lr": 7.817776490691442e-06, "train/step_canary_loss": 11.09375, "perf/step_duration_sec": 6.06073259213008, "perf/samples_per_sec": 5.609882878540018, "perf/tokens_per_sec": 3966.6821847935466, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 24041.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.8897891996016525, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561978.9685912, "event": "train_step", "step": 590, "epoch": 3, "metrics": {"train/step_loss": 2.323390669292874, "train/step_real_loss": 1.397017627954483, "train/lr": 7.502403599182098e-06, "train/step_canary_loss": 9.734375, "perf/step_duration_sec": 5.4169949509669095, "perf/samples_per_sec": 6.645751071555671, "perf/tokens_per_sec": 4397.456563209102, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 23821.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.9033228556059743, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774561993.300778, "event": "eval_step", "step": 590, "epoch": 3, "metrics": {"eval/loss": 0.8485451995987782, "eval/duration_sec": 14.319125022972003}}
{"timestamp": 1774562019.2149987, "event": "train_step", "step": 595, "epoch": 3, "metrics": {"train/step_loss": 2.3000132328755147, "train/step_real_loss": 1.2961090505123138, "train/lr": 7.22979683841816e-06, "train/step_canary_loss": 8.725, "perf/step_duration_sec": 6.022374538006261, "perf/samples_per_sec": 5.9777085886654255, "perf/tokens_per_sec": 4522.634689707783, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 27237.0, "perf/physical_batches": 6.0, "privacy/epsilon": 2.916756664733321, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774562045.2270596, "event": "train_step", "step": 600, "epoch": 3, "metrics": {"train/step_loss": 1.3016399614738696, "train/step_real_loss": 1.342316210269928, "train/lr": 7.0001481227643515e-06, "train/step_canary_loss": 0.0, "perf/step_duration_sec": 5.064759212080389, "perf/samples_per_sec": 6.318168082635413, "perf/tokens_per_sec": 5310.617716207647, "perf/logical_batch_size": 32.0, "perf/logical_token_count": 26897.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.9301904738606686, "system/cuda_memory_allocated_gb": 16.909874439239502, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774562060.5227625, "event": "eval_step", "step": 600, "epoch": 3, "metrics": {"eval/loss": 0.8485655058175325, "eval/duration_sec": 15.289162995060906}}
{"timestamp": 1774562085.4566998, "event": "train_step", "step": 605, "epoch": 3, "metrics": {"train/step_loss": 2.237206644482083, "train/step_real_loss": 1.2317012250423431, "train/lr": 6.813619124246933e-06, "train/step_canary_loss": 10.28125, "perf/step_duration_sec": 4.715751683106646, "perf/samples_per_sec": 7.6339897473744625, "perf/tokens_per_sec": 6242.695115915467, "perf/logical_batch_size": 36.0, "perf/logical_token_count": 29439.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.943624282988016, "system/cuda_memory_allocated_gb": 16.5317063331604, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774562111.906876, "event": "train_step", "step": 610, "epoch": 3, "metrics": {"train/step_loss": 2.055134582519531, "train/step_real_loss": 1.370850294828415, "train/lr": 6.670341158736986e-06, "train/step_canary_loss": 9.354166984558105, "perf/step_duration_sec": 5.474582785042003, "perf/samples_per_sec": 6.393181247643051, "perf/tokens_per_sec": 5057.006366885654, "perf/logical_batch_size": 35.0, "perf/logical_token_count": 27685.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.9569713855349518, "system/cuda_memory_allocated_gb": 16.436978340148926, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774562126.2327754, "event": "eval_step", "step": 610, "epoch": 3, "metrics": {"eval/loss": 0.8485312080857429, "eval/duration_sec": 14.309055346995592}}
{"timestamp": 1774562152.4836345, "event": "train_step", "step": 615, "epoch": 3, "metrics": {"train/step_loss": 1.89218209771549, "train/step_real_loss": 1.3737247288227081, "train/lr": 6.57041509350448e-06, "train/step_canary_loss": 10.1875, "perf/step_duration_sec": 5.486198005033657, "perf/samples_per_sec": 6.197370194951871, "perf/tokens_per_sec": 4890.818737380841, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 26832.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.9702172055473346, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774562178.1081915, "event": "train_step", "step": 620, "epoch": 3, "metrics": {"train/step_loss": 1.7868300606222713, "train/step_real_loss": 1.3467491269111633, "train/lr": 6.5139112762079875e-06, "train/step_canary_loss": 8.828125, "perf/step_duration_sec": 5.382014617091045, "perf/samples_per_sec": 6.317336986047959, "perf/tokens_per_sec": 4584.714415609806, "perf/logical_batch_size": 34.0, "perf/logical_token_count": 24675.0, "perf/physical_batches": 5.0, "privacy/epsilon": 2.983463025559717, "system/cuda_memory_allocated_gb": 16.34225034713745, "system/cuda_max_memory_allocated_gb": 74.68966627120972}}
{"timestamp": 1774562192.9885495, "event": "eval_step", "step": 620, "epoch": 3, "metrics": {"eval/loss": 0.8485483423891392, "eval/duration_sec": 14.864690768066794}}
{"timestamp": 1774562225.0871465, "event": "train_epoch", "step": 624, "epoch": 3, "metrics": {"train/epoch_loss": 2.120771038223466, "train/epoch_real_loss": 1.3435396100842179, "train/epoch_canary_loss": 9.79185872581857, "perf/epoch_duration_sec": 1383.5877013350837, "perf/epoch_samples_per_sec": 42.09852396331288, "perf/epoch_tokens_per_sec": 31900.168639407973, "perf/epoch_samples": 58247.0, "perf/epoch_tokens": 44136681.0, "system/cuda_epoch_peak_memory_gb": 74.68966627120972, "eval/loss": 0.8485401758754795, "eval/duration_sec": 14.38143059401773, "privacy/epsilon": 2.994059681569623}}
{"timestamp": 1774562234.1083977, "event": "audit_epoch", "step": 624, "epoch": 3, "metrics": {"audit/delta": 1e-05, "audit/num_canaries": 500.0, "audit/num_members": 250.0, "audit/paper_guess_fraction": 0.2, "audit/paper_guess_steps": 20.0, "audit/loss/auc": 0.528976, "audit/loss/empirical_epsilon/0.05": 0.08366204984486103, "audit/loss/empirical_epsilon/0.01": 0.0, "audit/loss/empirical_epsilon_details/0.05/epsilon": 0.08366204984486103, "audit/loss/empirical_epsilon_details/0.05/num_guesses": 95.0, "audit/loss/empirical_epsilon_details/0.05/correct_guesses": 58.0, "audit/loss/empirical_epsilon_details/0.01/epsilon": 0.0, "audit/loss/empirical_epsilon_details/0.01/num_guesses": 0.0, "audit/loss/empirical_epsilon_details/0.01/correct_guesses": 0.0, "audit/embedding/auc": 0.540056, "audit/embedding/empirical_epsilon/0.05": 0.0, "audit/embedding/empirical_epsilon/0.01": 0.0, "audit/embedding/empirical_epsilon_details/0.05/epsilon": 0.0, "audit/embedding/empirical_epsilon_details/0.05/num_guesses": 0.0, "audit/embedding/empirical_epsilon_details/0.05/correct_guesses": 0.0, "audit/embedding/empirical_epsilon_details/0.01/epsilon": 0.0, "audit/embedding/empirical_epsilon_details/0.01/num_guesses": 0.0, "audit/embedding/empirical_epsilon_details/0.01/correct_guesses": 0.0, "perf/audit_duration_sec": 5.839315079851076}}
{"timestamp": 1774562242.5995293, "event": "audit_final", "step": 624, "epoch": 3, "metrics": {"audit/delta": 1e-05, "audit/num_canaries": 500.0, "audit/num_members": 250.0, "audit/paper_guess_fraction": 0.2, "audit/paper_guess_steps": 20.0, "audit/loss/auc": 0.528976, "audit/loss/empirical_epsilon/0.05": 0.08366204984486103, "audit/loss/empirical_epsilon/0.01": 0.0, "audit/loss/empirical_epsilon_details/0.05/epsilon": 0.08366204984486103, "audit/loss/empirical_epsilon_details/0.05/num_guesses": 95.0, "audit/loss/empirical_epsilon_details/0.05/correct_guesses": 58.0, "audit/loss/empirical_epsilon_details/0.01/epsilon": 0.0, "audit/loss/empirical_epsilon_details/0.01/num_guesses": 0.0, "audit/loss/empirical_epsilon_details/0.01/correct_guesses": 0.0, "audit/embedding/auc": 0.540056, "audit/embedding/empirical_epsilon/0.05": 0.0, "audit/embedding/empirical_epsilon/0.01": 0.0, "audit/embedding/empirical_epsilon_details/0.05/epsilon": 0.0, "audit/embedding/empirical_epsilon_details/0.05/num_guesses": 0.0, "audit/embedding/empirical_epsilon_details/0.05/correct_guesses": 0.0, "audit/embedding/empirical_epsilon_details/0.01/epsilon": 0.0, "audit/embedding/empirical_epsilon_details/0.01/num_guesses": 0.0, "audit/embedding/empirical_epsilon_details/0.01/correct_guesses": 0.0}}
{"timestamp": 1774562243.1679335, "event": "energy_final", "step": 624, "epoch": null, "metrics": {"energy/codecarbon/duration": 4326.139577268856, "energy/codecarbon/emissions": 0.18796959736431926, "energy/codecarbon/emissions_rate": 4.3449730182535336e-05, "energy/codecarbon/cpu_power": 80.23442763916076, "energy/codecarbon/gpu_power": 4402.554613285673, "energy/codecarbon/ram_power": 38.0, "energy/codecarbon/cpu_energy": 0.09261863536672514, "energy/codecarbon/gpu_energy": 5.25827568300609, "energy/codecarbon/ram_energy": 0.043864955062854406, "energy/codecarbon/energy_consumed": 5.394759273435675, "energy/codecarbon/water_consumed": 0.0, "energy/codecarbon/cpu_count": 8.0, "energy/codecarbon/gpu_count": 8.0, "energy/codecarbon/longitude": 8.212, "energy/codecarbon/latitude": 47.4843, "energy/codecarbon/ram_total_size": 256.0, "energy/codecarbon/cpu_utilization_percent": 6.493086477618936, "energy/codecarbon/gpu_utilization_percent": 93.39248886805719, "energy/codecarbon/ram_utilization_percent": 26.041457698617293, "energy/codecarbon/ram_used_gb": 515.7987046457438, "energy/codecarbon/pue": 1.0, "energy/codecarbon/wue": 0.0}}