| { | |
| "best_metric": 1.5631914138793945, | |
| "best_model_checkpoint": "../res/Transformers/AST-ASVspoof5-Synthetic-Voice-Detection\\checkpoint-45590", | |
| "epoch": 3.0, | |
| "eval_steps": 500, | |
| "global_step": 68385, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.021934634788330774, | |
| "grad_norm": 0.0010007299715653062, | |
| "learning_rate": 4.963442275352782e-05, | |
| "loss": 0.0624, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.04386926957666155, | |
| "grad_norm": 0.0003039250732399523, | |
| "learning_rate": 4.926884550705564e-05, | |
| "loss": 0.0344, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.06580390436499232, | |
| "grad_norm": 0.0039045666344463825, | |
| "learning_rate": 4.890326826058346e-05, | |
| "loss": 0.0345, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.0877385391533231, | |
| "grad_norm": 0.00042319874046370387, | |
| "learning_rate": 4.853769101411129e-05, | |
| "loss": 0.0198, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.10967317394165386, | |
| "grad_norm": 0.0058428761549293995, | |
| "learning_rate": 4.817211376763911e-05, | |
| "loss": 0.0351, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.13160780872998465, | |
| "grad_norm": 0.0005889787571504712, | |
| "learning_rate": 4.780653652116692e-05, | |
| "loss": 0.0123, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.15354244351831542, | |
| "grad_norm": 0.0018479392165318131, | |
| "learning_rate": 4.7440959274694746e-05, | |
| "loss": 0.0135, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.1754770783066462, | |
| "grad_norm": 0.003831966780126095, | |
| "learning_rate": 4.7075382028222566e-05, | |
| "loss": 0.0195, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.19741171309497696, | |
| "grad_norm": 0.0023953751660883427, | |
| "learning_rate": 4.6709804781750386e-05, | |
| "loss": 0.0158, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.21934634788330773, | |
| "grad_norm": 0.0003263187827542424, | |
| "learning_rate": 4.6344227535278205e-05, | |
| "loss": 0.0173, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.24128098267163853, | |
| "grad_norm": 2.6989218895323575e-05, | |
| "learning_rate": 4.5978650288806025e-05, | |
| "loss": 0.0061, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.2632156174599693, | |
| "grad_norm": 0.00040426268242299557, | |
| "learning_rate": 4.561307304233385e-05, | |
| "loss": 0.019, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.28515025224830004, | |
| "grad_norm": 0.0005795055185444653, | |
| "learning_rate": 4.5247495795861664e-05, | |
| "loss": 0.013, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.30708488703663084, | |
| "grad_norm": 0.0019891748670488596, | |
| "learning_rate": 4.488191854938949e-05, | |
| "loss": 0.0098, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.32901952182496164, | |
| "grad_norm": 0.0010473368456587195, | |
| "learning_rate": 4.451634130291731e-05, | |
| "loss": 0.0274, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.3509541566132924, | |
| "grad_norm": 0.00016111508011817932, | |
| "learning_rate": 4.415076405644513e-05, | |
| "loss": 0.0056, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.3728887914016232, | |
| "grad_norm": 4.383545820019208e-05, | |
| "learning_rate": 4.378518680997295e-05, | |
| "loss": 0.0106, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.3948234261899539, | |
| "grad_norm": 0.0005566978361457586, | |
| "learning_rate": 4.341960956350077e-05, | |
| "loss": 0.0111, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.4167580609782847, | |
| "grad_norm": 0.002490273443982005, | |
| "learning_rate": 4.3054032317028595e-05, | |
| "loss": 0.0134, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.43869269576661546, | |
| "grad_norm": 0.012605056166648865, | |
| "learning_rate": 4.268845507055641e-05, | |
| "loss": 0.0106, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.46062733055494626, | |
| "grad_norm": 0.0012354288483038545, | |
| "learning_rate": 4.232287782408423e-05, | |
| "loss": 0.02, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.48256196534327706, | |
| "grad_norm": 1.0211567878723145, | |
| "learning_rate": 4.1957300577612054e-05, | |
| "loss": 0.0063, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.5044966001316078, | |
| "grad_norm": 0.0001449681876692921, | |
| "learning_rate": 4.159172333113987e-05, | |
| "loss": 0.0042, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 0.5264312349199386, | |
| "grad_norm": 7.361311872955412e-05, | |
| "learning_rate": 4.122614608466769e-05, | |
| "loss": 0.0, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.5483658697082694, | |
| "grad_norm": 0.0009549686801619828, | |
| "learning_rate": 4.086056883819551e-05, | |
| "loss": 0.0139, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 0.5703005044966001, | |
| "grad_norm": 0.0009273311588913202, | |
| "learning_rate": 4.049499159172334e-05, | |
| "loss": 0.0137, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 0.5922351392849309, | |
| "grad_norm": 0.0004449704720173031, | |
| "learning_rate": 4.012941434525115e-05, | |
| "loss": 0.0073, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 0.6141697740732617, | |
| "grad_norm": 3.233966344851069e-05, | |
| "learning_rate": 3.976383709877897e-05, | |
| "loss": 0.0, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 0.6361044088615925, | |
| "grad_norm": 0.000247443764237687, | |
| "learning_rate": 3.93982598523068e-05, | |
| "loss": 0.0107, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 0.6580390436499233, | |
| "grad_norm": 0.0010792565299198031, | |
| "learning_rate": 3.903268260583462e-05, | |
| "loss": 0.0064, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 0.679973678438254, | |
| "grad_norm": 0.0002366681583225727, | |
| "learning_rate": 3.866710535936244e-05, | |
| "loss": 0.0048, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 0.7019083132265848, | |
| "grad_norm": 0.00024113136169034988, | |
| "learning_rate": 3.8301528112890256e-05, | |
| "loss": 0.0087, | |
| "step": 16000 | |
| }, | |
| { | |
| "epoch": 0.7238429480149156, | |
| "grad_norm": 0.0059726969338953495, | |
| "learning_rate": 3.7935950866418076e-05, | |
| "loss": 0.0126, | |
| "step": 16500 | |
| }, | |
| { | |
| "epoch": 0.7457775828032464, | |
| "grad_norm": 0.00010310174548067153, | |
| "learning_rate": 3.7570373619945896e-05, | |
| "loss": 0.0027, | |
| "step": 17000 | |
| }, | |
| { | |
| "epoch": 0.7677122175915772, | |
| "grad_norm": 0.00020225635671522468, | |
| "learning_rate": 3.7204796373473715e-05, | |
| "loss": 0.015, | |
| "step": 17500 | |
| }, | |
| { | |
| "epoch": 0.7896468523799078, | |
| "grad_norm": 0.0002575612161308527, | |
| "learning_rate": 3.683921912700154e-05, | |
| "loss": 0.0038, | |
| "step": 18000 | |
| }, | |
| { | |
| "epoch": 0.8115814871682386, | |
| "grad_norm": 0.0002956839744001627, | |
| "learning_rate": 3.6473641880529354e-05, | |
| "loss": 0.0073, | |
| "step": 18500 | |
| }, | |
| { | |
| "epoch": 0.8335161219565694, | |
| "grad_norm": 0.0057256026193499565, | |
| "learning_rate": 3.6108064634057174e-05, | |
| "loss": 0.0124, | |
| "step": 19000 | |
| }, | |
| { | |
| "epoch": 0.8554507567449002, | |
| "grad_norm": 0.0009436457185074687, | |
| "learning_rate": 3.5742487387585e-05, | |
| "loss": 0.0055, | |
| "step": 19500 | |
| }, | |
| { | |
| "epoch": 0.8773853915332309, | |
| "grad_norm": 0.000185984157724306, | |
| "learning_rate": 3.537691014111282e-05, | |
| "loss": 0.0074, | |
| "step": 20000 | |
| }, | |
| { | |
| "epoch": 0.8993200263215617, | |
| "grad_norm": 0.0011451400350779295, | |
| "learning_rate": 3.501133289464064e-05, | |
| "loss": 0.0, | |
| "step": 20500 | |
| }, | |
| { | |
| "epoch": 0.9212546611098925, | |
| "grad_norm": 0.0002608870272524655, | |
| "learning_rate": 3.464575564816846e-05, | |
| "loss": 0.0054, | |
| "step": 21000 | |
| }, | |
| { | |
| "epoch": 0.9431892958982233, | |
| "grad_norm": 6.460679287556559e-05, | |
| "learning_rate": 3.428017840169628e-05, | |
| "loss": 0.0036, | |
| "step": 21500 | |
| }, | |
| { | |
| "epoch": 0.9651239306865541, | |
| "grad_norm": 0.0002977416734211147, | |
| "learning_rate": 3.39146011552241e-05, | |
| "loss": 0.0103, | |
| "step": 22000 | |
| }, | |
| { | |
| "epoch": 0.9870585654748848, | |
| "grad_norm": 8.106639870675281e-05, | |
| "learning_rate": 3.354902390875192e-05, | |
| "loss": 0.0042, | |
| "step": 22500 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.8470237672933665, | |
| "eval_f1": 0.8942219954670775, | |
| "eval_loss": 1.69542396068573, | |
| "eval_precision": 0.9672489546409696, | |
| "eval_recall": 0.831447963800905, | |
| "eval_runtime": 2882.9169, | |
| "eval_samples_per_second": 48.891, | |
| "eval_steps_per_second": 6.112, | |
| "step": 22795 | |
| }, | |
| { | |
| "epoch": 1.0089932002632156, | |
| "grad_norm": 4.19682874053251e-05, | |
| "learning_rate": 3.3183446662279744e-05, | |
| "loss": 0.0001, | |
| "step": 23000 | |
| }, | |
| { | |
| "epoch": 1.0309278350515463, | |
| "grad_norm": 5.8088691730517894e-05, | |
| "learning_rate": 3.2817869415807564e-05, | |
| "loss": 0.0, | |
| "step": 23500 | |
| }, | |
| { | |
| "epoch": 1.0528624698398772, | |
| "grad_norm": 0.0006424049497582018, | |
| "learning_rate": 3.245229216933538e-05, | |
| "loss": 0.005, | |
| "step": 24000 | |
| }, | |
| { | |
| "epoch": 1.0747971046282079, | |
| "grad_norm": 8.812173473415896e-05, | |
| "learning_rate": 3.20867149228632e-05, | |
| "loss": 0.0084, | |
| "step": 24500 | |
| }, | |
| { | |
| "epoch": 1.0967317394165388, | |
| "grad_norm": 0.0002326608228031546, | |
| "learning_rate": 3.172113767639102e-05, | |
| "loss": 0.0073, | |
| "step": 25000 | |
| }, | |
| { | |
| "epoch": 1.1186663742048695, | |
| "grad_norm": 0.00021554487466346473, | |
| "learning_rate": 3.135556042991884e-05, | |
| "loss": 0.0, | |
| "step": 25500 | |
| }, | |
| { | |
| "epoch": 1.1406010089932002, | |
| "grad_norm": 0.0071836900897324085, | |
| "learning_rate": 3.098998318344666e-05, | |
| "loss": 0.0131, | |
| "step": 26000 | |
| }, | |
| { | |
| "epoch": 1.162535643781531, | |
| "grad_norm": 0.00022567117412108928, | |
| "learning_rate": 3.062440593697448e-05, | |
| "loss": 0.0041, | |
| "step": 26500 | |
| }, | |
| { | |
| "epoch": 1.1844702785698618, | |
| "grad_norm": 0.00019994670583400875, | |
| "learning_rate": 3.0258828690502304e-05, | |
| "loss": 0.0023, | |
| "step": 27000 | |
| }, | |
| { | |
| "epoch": 1.2064049133581927, | |
| "grad_norm": 0.00010507569822948426, | |
| "learning_rate": 2.9893251444030124e-05, | |
| "loss": 0.0021, | |
| "step": 27500 | |
| }, | |
| { | |
| "epoch": 1.2283395481465234, | |
| "grad_norm": 0.00011653576075332239, | |
| "learning_rate": 2.9527674197557947e-05, | |
| "loss": 0.0028, | |
| "step": 28000 | |
| }, | |
| { | |
| "epoch": 1.250274182934854, | |
| "grad_norm": 0.000147501690662466, | |
| "learning_rate": 2.9162096951085767e-05, | |
| "loss": 0.0051, | |
| "step": 28500 | |
| }, | |
| { | |
| "epoch": 1.272208817723185, | |
| "grad_norm": 0.00015131481632124633, | |
| "learning_rate": 2.8796519704613583e-05, | |
| "loss": 0.0017, | |
| "step": 29000 | |
| }, | |
| { | |
| "epoch": 1.2941434525115156, | |
| "grad_norm": 6.996198499109596e-05, | |
| "learning_rate": 2.8430942458141406e-05, | |
| "loss": 0.0, | |
| "step": 29500 | |
| }, | |
| { | |
| "epoch": 1.3160780872998465, | |
| "grad_norm": 5.073639840702526e-05, | |
| "learning_rate": 2.8065365211669225e-05, | |
| "loss": 0.0005, | |
| "step": 30000 | |
| }, | |
| { | |
| "epoch": 1.3380127220881772, | |
| "grad_norm": 0.0008765258244238794, | |
| "learning_rate": 2.769978796519705e-05, | |
| "loss": 0.0034, | |
| "step": 30500 | |
| }, | |
| { | |
| "epoch": 1.359947356876508, | |
| "grad_norm": 0.0001492842857260257, | |
| "learning_rate": 2.7334210718724868e-05, | |
| "loss": 0.0, | |
| "step": 31000 | |
| }, | |
| { | |
| "epoch": 1.3818819916648388, | |
| "grad_norm": 7.198125240392983e-05, | |
| "learning_rate": 2.6968633472252684e-05, | |
| "loss": 0.0, | |
| "step": 31500 | |
| }, | |
| { | |
| "epoch": 1.4038166264531695, | |
| "grad_norm": 0.00025220203679054976, | |
| "learning_rate": 2.660305622578051e-05, | |
| "loss": 0.0099, | |
| "step": 32000 | |
| }, | |
| { | |
| "epoch": 1.4257512612415004, | |
| "grad_norm": 0.00011568542686291039, | |
| "learning_rate": 2.6237478979308327e-05, | |
| "loss": 0.0, | |
| "step": 32500 | |
| }, | |
| { | |
| "epoch": 1.447685896029831, | |
| "grad_norm": 0.00023261412570718676, | |
| "learning_rate": 2.587190173283615e-05, | |
| "loss": 0.0098, | |
| "step": 33000 | |
| }, | |
| { | |
| "epoch": 1.4696205308181618, | |
| "grad_norm": 0.0010702387662604451, | |
| "learning_rate": 2.550632448636397e-05, | |
| "loss": 0.0045, | |
| "step": 33500 | |
| }, | |
| { | |
| "epoch": 1.4915551656064927, | |
| "grad_norm": 0.0008456298382952809, | |
| "learning_rate": 2.514074723989179e-05, | |
| "loss": 0.0035, | |
| "step": 34000 | |
| }, | |
| { | |
| "epoch": 1.5134898003948234, | |
| "grad_norm": 0.0001115118502639234, | |
| "learning_rate": 2.4775169993419612e-05, | |
| "loss": 0.0021, | |
| "step": 34500 | |
| }, | |
| { | |
| "epoch": 1.5354244351831543, | |
| "grad_norm": 7.670137711102143e-05, | |
| "learning_rate": 2.440959274694743e-05, | |
| "loss": 0.0008, | |
| "step": 35000 | |
| }, | |
| { | |
| "epoch": 1.557359069971485, | |
| "grad_norm": 0.0001393236598232761, | |
| "learning_rate": 2.404401550047525e-05, | |
| "loss": 0.0044, | |
| "step": 35500 | |
| }, | |
| { | |
| "epoch": 1.5792937047598157, | |
| "grad_norm": 9.646616672398522e-05, | |
| "learning_rate": 2.367843825400307e-05, | |
| "loss": 0.0, | |
| "step": 36000 | |
| }, | |
| { | |
| "epoch": 1.6012283395481464, | |
| "grad_norm": 7.508257840527222e-05, | |
| "learning_rate": 2.3312861007530894e-05, | |
| "loss": 0.0, | |
| "step": 36500 | |
| }, | |
| { | |
| "epoch": 1.6231629743364773, | |
| "grad_norm": 6.9023335527163e-05, | |
| "learning_rate": 2.2947283761058713e-05, | |
| "loss": 0.0, | |
| "step": 37000 | |
| }, | |
| { | |
| "epoch": 1.6450976091248082, | |
| "grad_norm": 6.870364450151101e-05, | |
| "learning_rate": 2.2581706514586533e-05, | |
| "loss": 0.0, | |
| "step": 37500 | |
| }, | |
| { | |
| "epoch": 1.6670322439131389, | |
| "grad_norm": 10.176724433898926, | |
| "learning_rate": 2.2216129268114352e-05, | |
| "loss": 0.0053, | |
| "step": 38000 | |
| }, | |
| { | |
| "epoch": 1.6889668787014696, | |
| "grad_norm": 0.0005877121002413332, | |
| "learning_rate": 2.1850552021642172e-05, | |
| "loss": 0.0085, | |
| "step": 38500 | |
| }, | |
| { | |
| "epoch": 1.7109015134898002, | |
| "grad_norm": 0.0003466593334451318, | |
| "learning_rate": 2.1484974775169995e-05, | |
| "loss": 0.004, | |
| "step": 39000 | |
| }, | |
| { | |
| "epoch": 1.7328361482781311, | |
| "grad_norm": 12.523138046264648, | |
| "learning_rate": 2.1119397528697815e-05, | |
| "loss": 0.0028, | |
| "step": 39500 | |
| }, | |
| { | |
| "epoch": 1.754770783066462, | |
| "grad_norm": 0.00021529280638787895, | |
| "learning_rate": 2.0753820282225638e-05, | |
| "loss": 0.0007, | |
| "step": 40000 | |
| }, | |
| { | |
| "epoch": 1.7767054178547927, | |
| "grad_norm": 9.494409459875897e-05, | |
| "learning_rate": 2.0388243035753457e-05, | |
| "loss": 0.0, | |
| "step": 40500 | |
| }, | |
| { | |
| "epoch": 1.7986400526431234, | |
| "grad_norm": 8.512389467796311e-05, | |
| "learning_rate": 2.0022665789281277e-05, | |
| "loss": 0.0, | |
| "step": 41000 | |
| }, | |
| { | |
| "epoch": 1.8205746874314541, | |
| "grad_norm": 7.772703975206241e-05, | |
| "learning_rate": 1.9657088542809096e-05, | |
| "loss": 0.0, | |
| "step": 41500 | |
| }, | |
| { | |
| "epoch": 1.842509322219785, | |
| "grad_norm": 0.0005296979215927422, | |
| "learning_rate": 1.9291511296336916e-05, | |
| "loss": 0.0, | |
| "step": 42000 | |
| }, | |
| { | |
| "epoch": 1.864443957008116, | |
| "grad_norm": 3.345280856592581e-05, | |
| "learning_rate": 1.892593404986474e-05, | |
| "loss": 0.0, | |
| "step": 42500 | |
| }, | |
| { | |
| "epoch": 1.8863785917964466, | |
| "grad_norm": 2.8574026146088727e-05, | |
| "learning_rate": 1.856035680339256e-05, | |
| "loss": 0.0, | |
| "step": 43000 | |
| }, | |
| { | |
| "epoch": 1.9083132265847773, | |
| "grad_norm": 2.9599379558931105e-05, | |
| "learning_rate": 1.8194779556920378e-05, | |
| "loss": 0.0, | |
| "step": 43500 | |
| }, | |
| { | |
| "epoch": 1.930247861373108, | |
| "grad_norm": 2.845803101081401e-05, | |
| "learning_rate": 1.7829202310448198e-05, | |
| "loss": 0.0, | |
| "step": 44000 | |
| }, | |
| { | |
| "epoch": 1.952182496161439, | |
| "grad_norm": 2.3830825739423744e-05, | |
| "learning_rate": 1.7463625063976017e-05, | |
| "loss": 0.0, | |
| "step": 44500 | |
| }, | |
| { | |
| "epoch": 1.9741171309497698, | |
| "grad_norm": 1.564464400871657e-05, | |
| "learning_rate": 1.709804781750384e-05, | |
| "loss": 0.0, | |
| "step": 45000 | |
| }, | |
| { | |
| "epoch": 1.9960517657381005, | |
| "grad_norm": 1.654511652304791e-05, | |
| "learning_rate": 1.673247057103166e-05, | |
| "loss": 0.0, | |
| "step": 45500 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8489322454771195, | |
| "eval_f1": 0.9013532483055441, | |
| "eval_loss": 1.5631914138793945, | |
| "eval_precision": 0.9156963336000377, | |
| "eval_recall": 0.8874525616698292, | |
| "eval_runtime": 2678.5689, | |
| "eval_samples_per_second": 52.621, | |
| "eval_steps_per_second": 6.578, | |
| "step": 45590 | |
| }, | |
| { | |
| "epoch": 2.017986400526431, | |
| "grad_norm": 1.1972474567301106e-05, | |
| "learning_rate": 1.636689332455948e-05, | |
| "loss": 0.0, | |
| "step": 46000 | |
| }, | |
| { | |
| "epoch": 2.039921035314762, | |
| "grad_norm": 2.4531369490432553e-05, | |
| "learning_rate": 1.60013160780873e-05, | |
| "loss": 0.0062, | |
| "step": 46500 | |
| }, | |
| { | |
| "epoch": 2.0618556701030926, | |
| "grad_norm": 7.93306899140589e-05, | |
| "learning_rate": 1.5635738831615122e-05, | |
| "loss": 0.0038, | |
| "step": 47000 | |
| }, | |
| { | |
| "epoch": 2.0837903048914237, | |
| "grad_norm": 2.241317451989744e-05, | |
| "learning_rate": 1.5270161585142942e-05, | |
| "loss": 0.0, | |
| "step": 47500 | |
| }, | |
| { | |
| "epoch": 2.1057249396797544, | |
| "grad_norm": 1.646195050852839e-05, | |
| "learning_rate": 1.4904584338670763e-05, | |
| "loss": 0.0, | |
| "step": 48000 | |
| }, | |
| { | |
| "epoch": 2.127659574468085, | |
| "grad_norm": 2.041865991486702e-05, | |
| "learning_rate": 1.4539007092198581e-05, | |
| "loss": 0.0, | |
| "step": 48500 | |
| }, | |
| { | |
| "epoch": 2.1495942092564158, | |
| "grad_norm": 1.3043787475908175e-05, | |
| "learning_rate": 1.4173429845726402e-05, | |
| "loss": 0.0031, | |
| "step": 49000 | |
| }, | |
| { | |
| "epoch": 2.1715288440447464, | |
| "grad_norm": 1.759601764206309e-05, | |
| "learning_rate": 1.3807852599254223e-05, | |
| "loss": 0.0, | |
| "step": 49500 | |
| }, | |
| { | |
| "epoch": 2.1934634788330776, | |
| "grad_norm": 9.869835594145115e-06, | |
| "learning_rate": 1.3442275352782043e-05, | |
| "loss": 0.0, | |
| "step": 50000 | |
| }, | |
| { | |
| "epoch": 2.2153981136214083, | |
| "grad_norm": 1.1258347512921318e-05, | |
| "learning_rate": 1.3076698106309864e-05, | |
| "loss": 0.0019, | |
| "step": 50500 | |
| }, | |
| { | |
| "epoch": 2.237332748409739, | |
| "grad_norm": 9.829653572523966e-05, | |
| "learning_rate": 1.2711120859837682e-05, | |
| "loss": 0.0, | |
| "step": 51000 | |
| }, | |
| { | |
| "epoch": 2.2592673831980696, | |
| "grad_norm": 6.886749906698242e-05, | |
| "learning_rate": 1.2345543613365505e-05, | |
| "loss": 0.0, | |
| "step": 51500 | |
| }, | |
| { | |
| "epoch": 2.2812020179864003, | |
| "grad_norm": 1.0710909918998368e-05, | |
| "learning_rate": 1.1979966366893325e-05, | |
| "loss": 0.0, | |
| "step": 52000 | |
| }, | |
| { | |
| "epoch": 2.3031366527747315, | |
| "grad_norm": 6.866812327643856e-06, | |
| "learning_rate": 1.1614389120421146e-05, | |
| "loss": 0.0, | |
| "step": 52500 | |
| }, | |
| { | |
| "epoch": 2.325071287563062, | |
| "grad_norm": 3.13724776788149e-05, | |
| "learning_rate": 1.1248811873948966e-05, | |
| "loss": 0.0015, | |
| "step": 53000 | |
| }, | |
| { | |
| "epoch": 2.347005922351393, | |
| "grad_norm": 1.3903067156206816e-05, | |
| "learning_rate": 1.0883234627476787e-05, | |
| "loss": 0.0015, | |
| "step": 53500 | |
| }, | |
| { | |
| "epoch": 2.3689405571397235, | |
| "grad_norm": 9.283921826863661e-06, | |
| "learning_rate": 1.0517657381004607e-05, | |
| "loss": 0.0, | |
| "step": 54000 | |
| }, | |
| { | |
| "epoch": 2.390875191928054, | |
| "grad_norm": 9.730028068588581e-06, | |
| "learning_rate": 1.0152080134532426e-05, | |
| "loss": 0.0017, | |
| "step": 54500 | |
| }, | |
| { | |
| "epoch": 2.4128098267163853, | |
| "grad_norm": 0.0004968822468072176, | |
| "learning_rate": 9.786502888060247e-06, | |
| "loss": 0.0, | |
| "step": 55000 | |
| }, | |
| { | |
| "epoch": 2.434744461504716, | |
| "grad_norm": 1.0726791515480727e-05, | |
| "learning_rate": 9.420925641588069e-06, | |
| "loss": 0.0, | |
| "step": 55500 | |
| }, | |
| { | |
| "epoch": 2.4566790962930467, | |
| "grad_norm": 9.191048775392119e-06, | |
| "learning_rate": 9.055348395115888e-06, | |
| "loss": 0.0, | |
| "step": 56000 | |
| }, | |
| { | |
| "epoch": 2.4786137310813774, | |
| "grad_norm": 6.658553957095137e-06, | |
| "learning_rate": 8.68977114864371e-06, | |
| "loss": 0.0, | |
| "step": 56500 | |
| }, | |
| { | |
| "epoch": 2.500548365869708, | |
| "grad_norm": 5.824945674248738e-06, | |
| "learning_rate": 8.32419390217153e-06, | |
| "loss": 0.0, | |
| "step": 57000 | |
| }, | |
| { | |
| "epoch": 2.522483000658039, | |
| "grad_norm": 6.866499916213797e-06, | |
| "learning_rate": 7.95861665569935e-06, | |
| "loss": 0.0, | |
| "step": 57500 | |
| }, | |
| { | |
| "epoch": 2.54441763544637, | |
| "grad_norm": 5.9728527048719116e-06, | |
| "learning_rate": 7.593039409227171e-06, | |
| "loss": 0.0, | |
| "step": 58000 | |
| }, | |
| { | |
| "epoch": 2.5663522702347006, | |
| "grad_norm": 5.160507498658262e-06, | |
| "learning_rate": 7.2274621627549905e-06, | |
| "loss": 0.0, | |
| "step": 58500 | |
| }, | |
| { | |
| "epoch": 2.5882869050230313, | |
| "grad_norm": 4.935933702654438e-06, | |
| "learning_rate": 6.861884916282811e-06, | |
| "loss": 0.0, | |
| "step": 59000 | |
| }, | |
| { | |
| "epoch": 2.610221539811362, | |
| "grad_norm": 6.009084700053791e-06, | |
| "learning_rate": 6.496307669810631e-06, | |
| "loss": 0.0, | |
| "step": 59500 | |
| }, | |
| { | |
| "epoch": 2.632156174599693, | |
| "grad_norm": 5.124590188643197e-06, | |
| "learning_rate": 6.130730423338452e-06, | |
| "loss": 0.0, | |
| "step": 60000 | |
| }, | |
| { | |
| "epoch": 2.6540908093880238, | |
| "grad_norm": 4.586262548400555e-06, | |
| "learning_rate": 5.7651531768662715e-06, | |
| "loss": 0.0, | |
| "step": 60500 | |
| }, | |
| { | |
| "epoch": 2.6760254441763545, | |
| "grad_norm": 4.370785063656513e-06, | |
| "learning_rate": 5.399575930394093e-06, | |
| "loss": 0.0, | |
| "step": 61000 | |
| }, | |
| { | |
| "epoch": 2.697960078964685, | |
| "grad_norm": 2.011685182878864e-06, | |
| "learning_rate": 5.033998683921913e-06, | |
| "loss": 0.0, | |
| "step": 61500 | |
| }, | |
| { | |
| "epoch": 2.719894713753016, | |
| "grad_norm": 2.3848970158724114e-06, | |
| "learning_rate": 4.668421437449734e-06, | |
| "loss": 0.0, | |
| "step": 62000 | |
| }, | |
| { | |
| "epoch": 2.741829348541347, | |
| "grad_norm": 1.7811821635405067e-06, | |
| "learning_rate": 4.302844190977553e-06, | |
| "loss": 0.0, | |
| "step": 62500 | |
| }, | |
| { | |
| "epoch": 2.7637639833296777, | |
| "grad_norm": 1.9688261545525165e-06, | |
| "learning_rate": 3.9372669445053745e-06, | |
| "loss": 0.0, | |
| "step": 63000 | |
| }, | |
| { | |
| "epoch": 2.7856986181180083, | |
| "grad_norm": 1.5722528132755542e-06, | |
| "learning_rate": 3.5716896980331945e-06, | |
| "loss": 0.0, | |
| "step": 63500 | |
| }, | |
| { | |
| "epoch": 2.807633252906339, | |
| "grad_norm": 1.144167868005752e-06, | |
| "learning_rate": 3.206112451561015e-06, | |
| "loss": 0.0, | |
| "step": 64000 | |
| }, | |
| { | |
| "epoch": 2.8295678876946697, | |
| "grad_norm": 1.0192333093073103e-06, | |
| "learning_rate": 2.8405352050888354e-06, | |
| "loss": 0.0, | |
| "step": 64500 | |
| }, | |
| { | |
| "epoch": 2.851502522483001, | |
| "grad_norm": 1.021494654196431e-06, | |
| "learning_rate": 2.474957958616656e-06, | |
| "loss": 0.0, | |
| "step": 65000 | |
| }, | |
| { | |
| "epoch": 2.8734371572713315, | |
| "grad_norm": 1.6064728924902738e-06, | |
| "learning_rate": 2.109380712144476e-06, | |
| "loss": 0.0, | |
| "step": 65500 | |
| }, | |
| { | |
| "epoch": 2.895371792059662, | |
| "grad_norm": 1.136414766733651e-06, | |
| "learning_rate": 1.7438034656722967e-06, | |
| "loss": 0.0, | |
| "step": 66000 | |
| }, | |
| { | |
| "epoch": 2.917306426847993, | |
| "grad_norm": 1.4771685528103262e-06, | |
| "learning_rate": 1.3782262192001172e-06, | |
| "loss": 0.0, | |
| "step": 66500 | |
| }, | |
| { | |
| "epoch": 2.9392410616363236, | |
| "grad_norm": 6.969395940359391e-07, | |
| "learning_rate": 1.0126489727279376e-06, | |
| "loss": 0.0, | |
| "step": 67000 | |
| }, | |
| { | |
| "epoch": 2.9611756964246547, | |
| "grad_norm": 6.290829333011061e-07, | |
| "learning_rate": 6.470717262557578e-07, | |
| "loss": 0.0, | |
| "step": 67500 | |
| }, | |
| { | |
| "epoch": 2.9831103312129854, | |
| "grad_norm": 6.513329253721167e-07, | |
| "learning_rate": 2.814944797835783e-07, | |
| "loss": 0.0, | |
| "step": 68000 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8333451578573963, | |
| "eval_f1": 0.8891604695934469, | |
| "eval_loss": 2.2820544242858887, | |
| "eval_precision": 0.9208988192978341, | |
| "eval_recall": 0.8595369289154868, | |
| "eval_runtime": 2698.6211, | |
| "eval_samples_per_second": 52.23, | |
| "eval_steps_per_second": 6.529, | |
| "step": 68385 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "step": 68385, | |
| "total_flos": 3.708202600365923e+19, | |
| "train_loss": 0.005184370860593931, | |
| "train_runtime": 29657.93, | |
| "train_samples_per_second": 18.446, | |
| "train_steps_per_second": 2.306 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 68385, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 3, | |
| "save_steps": 500, | |
| "stateful_callbacks": { | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": true | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 3.708202600365923e+19, | |
| "train_batch_size": 8, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |