| { | |
| "best_metric": 1.4281049966812134, | |
| "best_model_checkpoint": "../res/Transformers/VIT-ASVspoof5-ConstantQ-Synthetic-Voice-Detection\\checkpoint-22795", | |
| "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": 0.2798495888710022, | |
| "learning_rate": 4.963442275352782e-05, | |
| "loss": 0.183, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.04386926957666155, | |
| "grad_norm": 0.06101477891206741, | |
| "learning_rate": 4.926884550705564e-05, | |
| "loss": 0.0997, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.06580390436499232, | |
| "grad_norm": 0.09287368506193161, | |
| "learning_rate": 4.890326826058346e-05, | |
| "loss": 0.1086, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.0877385391533231, | |
| "grad_norm": 0.015171859413385391, | |
| "learning_rate": 4.853769101411129e-05, | |
| "loss": 0.0726, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.10967317394165386, | |
| "grad_norm": 0.018528642132878304, | |
| "learning_rate": 4.817211376763911e-05, | |
| "loss": 0.0818, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.13160780872998465, | |
| "grad_norm": 0.2906191349029541, | |
| "learning_rate": 4.780653652116692e-05, | |
| "loss": 0.0657, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.15354244351831542, | |
| "grad_norm": 0.15548314154148102, | |
| "learning_rate": 4.7440959274694746e-05, | |
| "loss": 0.0786, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.1754770783066462, | |
| "grad_norm": 0.0061602904461324215, | |
| "learning_rate": 4.7075382028222566e-05, | |
| "loss": 0.0466, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.19741171309497696, | |
| "grad_norm": 0.04430825263261795, | |
| "learning_rate": 4.6709804781750386e-05, | |
| "loss": 0.0556, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.21934634788330773, | |
| "grad_norm": 0.2562808394432068, | |
| "learning_rate": 4.6344227535278205e-05, | |
| "loss": 0.0482, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.24128098267163853, | |
| "grad_norm": 0.02834293432533741, | |
| "learning_rate": 4.5978650288806025e-05, | |
| "loss": 0.0598, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.2632156174599693, | |
| "grad_norm": 0.014705098234117031, | |
| "learning_rate": 4.561307304233385e-05, | |
| "loss": 0.0574, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.28515025224830004, | |
| "grad_norm": 4.810993194580078, | |
| "learning_rate": 4.5247495795861664e-05, | |
| "loss": 0.0497, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.30708488703663084, | |
| "grad_norm": 0.021046284586191177, | |
| "learning_rate": 4.488191854938949e-05, | |
| "loss": 0.0493, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.32901952182496164, | |
| "grad_norm": 0.0069536808878183365, | |
| "learning_rate": 4.451634130291731e-05, | |
| "loss": 0.0551, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.3509541566132924, | |
| "grad_norm": 0.20052477717399597, | |
| "learning_rate": 4.415076405644513e-05, | |
| "loss": 0.0414, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.3728887914016232, | |
| "grad_norm": 0.0256254430860281, | |
| "learning_rate": 4.378518680997295e-05, | |
| "loss": 0.0534, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.3948234261899539, | |
| "grad_norm": 0.051042962819337845, | |
| "learning_rate": 4.341960956350077e-05, | |
| "loss": 0.0391, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.4167580609782847, | |
| "grad_norm": 0.002523040631785989, | |
| "learning_rate": 4.3054032317028595e-05, | |
| "loss": 0.0391, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.43869269576661546, | |
| "grad_norm": 0.22332483530044556, | |
| "learning_rate": 4.268845507055641e-05, | |
| "loss": 0.0403, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.46062733055494626, | |
| "grad_norm": 0.007270669564604759, | |
| "learning_rate": 4.232287782408423e-05, | |
| "loss": 0.0368, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.48256196534327706, | |
| "grad_norm": 0.008023562841117382, | |
| "learning_rate": 4.1957300577612054e-05, | |
| "loss": 0.0409, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.5044966001316078, | |
| "grad_norm": 0.006728711538016796, | |
| "learning_rate": 4.159172333113987e-05, | |
| "loss": 0.0365, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 0.5264312349199386, | |
| "grad_norm": 0.016508929431438446, | |
| "learning_rate": 4.122614608466769e-05, | |
| "loss": 0.035, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.5483658697082694, | |
| "grad_norm": 0.10062957555055618, | |
| "learning_rate": 4.086056883819551e-05, | |
| "loss": 0.037, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 0.5703005044966001, | |
| "grad_norm": 0.27828025817871094, | |
| "learning_rate": 4.049499159172334e-05, | |
| "loss": 0.03, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 0.5922351392849309, | |
| "grad_norm": 0.11781612783670425, | |
| "learning_rate": 4.012941434525115e-05, | |
| "loss": 0.0319, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 0.6141697740732617, | |
| "grad_norm": 0.007026828359812498, | |
| "learning_rate": 3.976383709877897e-05, | |
| "loss": 0.046, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 0.6361044088615925, | |
| "grad_norm": 0.0018289884319528937, | |
| "learning_rate": 3.93982598523068e-05, | |
| "loss": 0.0252, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 0.6580390436499233, | |
| "grad_norm": 0.005449921824038029, | |
| "learning_rate": 3.903268260583462e-05, | |
| "loss": 0.0357, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 0.679973678438254, | |
| "grad_norm": 0.006275506690144539, | |
| "learning_rate": 3.866710535936244e-05, | |
| "loss": 0.0343, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 0.7019083132265848, | |
| "grad_norm": 0.007306075654923916, | |
| "learning_rate": 3.8301528112890256e-05, | |
| "loss": 0.0299, | |
| "step": 16000 | |
| }, | |
| { | |
| "epoch": 0.7238429480149156, | |
| "grad_norm": 0.004079036880284548, | |
| "learning_rate": 3.7935950866418076e-05, | |
| "loss": 0.0303, | |
| "step": 16500 | |
| }, | |
| { | |
| "epoch": 0.7457775828032464, | |
| "grad_norm": 0.0035325633361935616, | |
| "learning_rate": 3.7570373619945896e-05, | |
| "loss": 0.03, | |
| "step": 17000 | |
| }, | |
| { | |
| "epoch": 0.7677122175915772, | |
| "grad_norm": 0.020647136494517326, | |
| "learning_rate": 3.7204796373473715e-05, | |
| "loss": 0.0296, | |
| "step": 17500 | |
| }, | |
| { | |
| "epoch": 0.7896468523799078, | |
| "grad_norm": 0.004098346456885338, | |
| "learning_rate": 3.683921912700154e-05, | |
| "loss": 0.0188, | |
| "step": 18000 | |
| }, | |
| { | |
| "epoch": 0.8115814871682386, | |
| "grad_norm": 0.0028982735238969326, | |
| "learning_rate": 3.6473641880529354e-05, | |
| "loss": 0.0215, | |
| "step": 18500 | |
| }, | |
| { | |
| "epoch": 0.8335161219565694, | |
| "grad_norm": 0.006535276770591736, | |
| "learning_rate": 3.6108064634057174e-05, | |
| "loss": 0.0296, | |
| "step": 19000 | |
| }, | |
| { | |
| "epoch": 0.8554507567449002, | |
| "grad_norm": 0.017595956102013588, | |
| "learning_rate": 3.5742487387585e-05, | |
| "loss": 0.0323, | |
| "step": 19500 | |
| }, | |
| { | |
| "epoch": 0.8773853915332309, | |
| "grad_norm": 0.014820819720625877, | |
| "learning_rate": 3.537691014111282e-05, | |
| "loss": 0.0216, | |
| "step": 20000 | |
| }, | |
| { | |
| "epoch": 0.8993200263215617, | |
| "grad_norm": 0.0010706137400120497, | |
| "learning_rate": 3.501133289464064e-05, | |
| "loss": 0.0239, | |
| "step": 20500 | |
| }, | |
| { | |
| "epoch": 0.9212546611098925, | |
| "grad_norm": 0.0014843459939584136, | |
| "learning_rate": 3.464575564816846e-05, | |
| "loss": 0.0201, | |
| "step": 21000 | |
| }, | |
| { | |
| "epoch": 0.9431892958982233, | |
| "grad_norm": 0.0009596743038855493, | |
| "learning_rate": 3.428017840169628e-05, | |
| "loss": 0.0241, | |
| "step": 21500 | |
| }, | |
| { | |
| "epoch": 0.9651239306865541, | |
| "grad_norm": 0.019483808428049088, | |
| "learning_rate": 3.39146011552241e-05, | |
| "loss": 0.0292, | |
| "step": 22000 | |
| }, | |
| { | |
| "epoch": 0.9870585654748848, | |
| "grad_norm": 0.0010368123184889555, | |
| "learning_rate": 3.354902390875192e-05, | |
| "loss": 0.0145, | |
| "step": 22500 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.7289039393972712, | |
| "eval_f1": 0.8193136601129977, | |
| "eval_loss": 1.4281049966812134, | |
| "eval_precision": 0.7274868564884701, | |
| "eval_recall": 0.9376708767339271, | |
| "eval_runtime": 581.2604, | |
| "eval_samples_per_second": 156.475, | |
| "eval_steps_per_second": 19.561, | |
| "step": 22795 | |
| }, | |
| { | |
| "epoch": 1.0089932002632156, | |
| "grad_norm": 0.35828471183776855, | |
| "learning_rate": 3.3183446662279744e-05, | |
| "loss": 0.0272, | |
| "step": 23000 | |
| }, | |
| { | |
| "epoch": 1.0309278350515463, | |
| "grad_norm": 0.0029607457108795643, | |
| "learning_rate": 3.2817869415807564e-05, | |
| "loss": 0.0196, | |
| "step": 23500 | |
| }, | |
| { | |
| "epoch": 1.0528624698398772, | |
| "grad_norm": 0.00152340077329427, | |
| "learning_rate": 3.245229216933538e-05, | |
| "loss": 0.0166, | |
| "step": 24000 | |
| }, | |
| { | |
| "epoch": 1.0747971046282079, | |
| "grad_norm": 0.0016208962770178914, | |
| "learning_rate": 3.20867149228632e-05, | |
| "loss": 0.0284, | |
| "step": 24500 | |
| }, | |
| { | |
| "epoch": 1.0967317394165388, | |
| "grad_norm": 0.0006961998296901584, | |
| "learning_rate": 3.172113767639102e-05, | |
| "loss": 0.0138, | |
| "step": 25000 | |
| }, | |
| { | |
| "epoch": 1.1186663742048695, | |
| "grad_norm": 0.0017152460059151053, | |
| "learning_rate": 3.135556042991884e-05, | |
| "loss": 0.0166, | |
| "step": 25500 | |
| }, | |
| { | |
| "epoch": 1.1406010089932002, | |
| "grad_norm": 0.0005695988656952977, | |
| "learning_rate": 3.098998318344666e-05, | |
| "loss": 0.0141, | |
| "step": 26000 | |
| }, | |
| { | |
| "epoch": 1.162535643781531, | |
| "grad_norm": 0.0022228530142456293, | |
| "learning_rate": 3.062440593697448e-05, | |
| "loss": 0.0162, | |
| "step": 26500 | |
| }, | |
| { | |
| "epoch": 1.1844702785698618, | |
| "grad_norm": 0.0008670862298458815, | |
| "learning_rate": 3.0258828690502304e-05, | |
| "loss": 0.0142, | |
| "step": 27000 | |
| }, | |
| { | |
| "epoch": 1.2064049133581927, | |
| "grad_norm": 0.01853255368769169, | |
| "learning_rate": 2.9893251444030124e-05, | |
| "loss": 0.0152, | |
| "step": 27500 | |
| }, | |
| { | |
| "epoch": 1.2283395481465234, | |
| "grad_norm": 0.0005061990232206881, | |
| "learning_rate": 2.9527674197557947e-05, | |
| "loss": 0.0156, | |
| "step": 28000 | |
| }, | |
| { | |
| "epoch": 1.250274182934854, | |
| "grad_norm": 0.0004909224808216095, | |
| "learning_rate": 2.9162096951085767e-05, | |
| "loss": 0.016, | |
| "step": 28500 | |
| }, | |
| { | |
| "epoch": 1.272208817723185, | |
| "grad_norm": 0.0007474181475117803, | |
| "learning_rate": 2.8796519704613583e-05, | |
| "loss": 0.015, | |
| "step": 29000 | |
| }, | |
| { | |
| "epoch": 1.2941434525115156, | |
| "grad_norm": 0.005266433581709862, | |
| "learning_rate": 2.8430942458141406e-05, | |
| "loss": 0.0205, | |
| "step": 29500 | |
| }, | |
| { | |
| "epoch": 1.3160780872998465, | |
| "grad_norm": 0.00045672807027585804, | |
| "learning_rate": 2.8065365211669225e-05, | |
| "loss": 0.0032, | |
| "step": 30000 | |
| }, | |
| { | |
| "epoch": 1.3380127220881772, | |
| "grad_norm": 0.000811495934613049, | |
| "learning_rate": 2.769978796519705e-05, | |
| "loss": 0.0158, | |
| "step": 30500 | |
| }, | |
| { | |
| "epoch": 1.359947356876508, | |
| "grad_norm": 0.06001759320497513, | |
| "learning_rate": 2.7334210718724868e-05, | |
| "loss": 0.0182, | |
| "step": 31000 | |
| }, | |
| { | |
| "epoch": 1.3818819916648388, | |
| "grad_norm": 0.0009237139020115137, | |
| "learning_rate": 2.6968633472252684e-05, | |
| "loss": 0.0089, | |
| "step": 31500 | |
| }, | |
| { | |
| "epoch": 1.4038166264531695, | |
| "grad_norm": 0.001214427873492241, | |
| "learning_rate": 2.660305622578051e-05, | |
| "loss": 0.0095, | |
| "step": 32000 | |
| }, | |
| { | |
| "epoch": 1.4257512612415004, | |
| "grad_norm": 0.0005619340226985514, | |
| "learning_rate": 2.6237478979308327e-05, | |
| "loss": 0.0123, | |
| "step": 32500 | |
| }, | |
| { | |
| "epoch": 1.447685896029831, | |
| "grad_norm": 0.00035489199217408895, | |
| "learning_rate": 2.587190173283615e-05, | |
| "loss": 0.0149, | |
| "step": 33000 | |
| }, | |
| { | |
| "epoch": 1.4696205308181618, | |
| "grad_norm": 0.0005062100244686007, | |
| "learning_rate": 2.550632448636397e-05, | |
| "loss": 0.0151, | |
| "step": 33500 | |
| }, | |
| { | |
| "epoch": 1.4915551656064927, | |
| "grad_norm": 0.00024328318249899894, | |
| "learning_rate": 2.514074723989179e-05, | |
| "loss": 0.0046, | |
| "step": 34000 | |
| }, | |
| { | |
| "epoch": 1.5134898003948234, | |
| "grad_norm": 0.0002546838077250868, | |
| "learning_rate": 2.4775169993419612e-05, | |
| "loss": 0.0145, | |
| "step": 34500 | |
| }, | |
| { | |
| "epoch": 1.5354244351831543, | |
| "grad_norm": 0.0003621385258156806, | |
| "learning_rate": 2.440959274694743e-05, | |
| "loss": 0.0096, | |
| "step": 35000 | |
| }, | |
| { | |
| "epoch": 1.557359069971485, | |
| "grad_norm": 0.001115673454478383, | |
| "learning_rate": 2.404401550047525e-05, | |
| "loss": 0.0125, | |
| "step": 35500 | |
| }, | |
| { | |
| "epoch": 1.5792937047598157, | |
| "grad_norm": 0.0003813185903709382, | |
| "learning_rate": 2.367843825400307e-05, | |
| "loss": 0.0162, | |
| "step": 36000 | |
| }, | |
| { | |
| "epoch": 1.6012283395481464, | |
| "grad_norm": 0.000716285954695195, | |
| "learning_rate": 2.3312861007530894e-05, | |
| "loss": 0.0138, | |
| "step": 36500 | |
| }, | |
| { | |
| "epoch": 1.6231629743364773, | |
| "grad_norm": 0.0005207060021348298, | |
| "learning_rate": 2.2947283761058713e-05, | |
| "loss": 0.0083, | |
| "step": 37000 | |
| }, | |
| { | |
| "epoch": 1.6450976091248082, | |
| "grad_norm": 0.05391339957714081, | |
| "learning_rate": 2.2581706514586533e-05, | |
| "loss": 0.0139, | |
| "step": 37500 | |
| }, | |
| { | |
| "epoch": 1.6670322439131389, | |
| "grad_norm": 0.00044836971210315824, | |
| "learning_rate": 2.2216129268114352e-05, | |
| "loss": 0.0098, | |
| "step": 38000 | |
| }, | |
| { | |
| "epoch": 1.6889668787014696, | |
| "grad_norm": 0.0029425013344734907, | |
| "learning_rate": 2.1850552021642172e-05, | |
| "loss": 0.0115, | |
| "step": 38500 | |
| }, | |
| { | |
| "epoch": 1.7109015134898002, | |
| "grad_norm": 0.0010432926937937737, | |
| "learning_rate": 2.1484974775169995e-05, | |
| "loss": 0.0106, | |
| "step": 39000 | |
| }, | |
| { | |
| "epoch": 1.7328361482781311, | |
| "grad_norm": 0.036389946937561035, | |
| "learning_rate": 2.1119397528697815e-05, | |
| "loss": 0.01, | |
| "step": 39500 | |
| }, | |
| { | |
| "epoch": 1.754770783066462, | |
| "grad_norm": 0.08681457489728928, | |
| "learning_rate": 2.0753820282225638e-05, | |
| "loss": 0.0107, | |
| "step": 40000 | |
| }, | |
| { | |
| "epoch": 1.7767054178547927, | |
| "grad_norm": 0.0002747771213762462, | |
| "learning_rate": 2.0388243035753457e-05, | |
| "loss": 0.0132, | |
| "step": 40500 | |
| }, | |
| { | |
| "epoch": 1.7986400526431234, | |
| "grad_norm": 0.00030470959609374404, | |
| "learning_rate": 2.0022665789281277e-05, | |
| "loss": 0.0068, | |
| "step": 41000 | |
| }, | |
| { | |
| "epoch": 1.8205746874314541, | |
| "grad_norm": 0.006674638018012047, | |
| "learning_rate": 1.9657088542809096e-05, | |
| "loss": 0.0116, | |
| "step": 41500 | |
| }, | |
| { | |
| "epoch": 1.842509322219785, | |
| "grad_norm": 0.024995682761073112, | |
| "learning_rate": 1.9291511296336916e-05, | |
| "loss": 0.0042, | |
| "step": 42000 | |
| }, | |
| { | |
| "epoch": 1.864443957008116, | |
| "grad_norm": 0.0002595706027932465, | |
| "learning_rate": 1.892593404986474e-05, | |
| "loss": 0.0147, | |
| "step": 42500 | |
| }, | |
| { | |
| "epoch": 1.8863785917964466, | |
| "grad_norm": 0.00032524587004445493, | |
| "learning_rate": 1.856035680339256e-05, | |
| "loss": 0.0126, | |
| "step": 43000 | |
| }, | |
| { | |
| "epoch": 1.9083132265847773, | |
| "grad_norm": 0.004932080861181021, | |
| "learning_rate": 1.8194779556920378e-05, | |
| "loss": 0.006, | |
| "step": 43500 | |
| }, | |
| { | |
| "epoch": 1.930247861373108, | |
| "grad_norm": 0.0003894896653946489, | |
| "learning_rate": 1.7829202310448198e-05, | |
| "loss": 0.011, | |
| "step": 44000 | |
| }, | |
| { | |
| "epoch": 1.952182496161439, | |
| "grad_norm": 0.0002865718270186335, | |
| "learning_rate": 1.7463625063976017e-05, | |
| "loss": 0.0046, | |
| "step": 44500 | |
| }, | |
| { | |
| "epoch": 1.9741171309497698, | |
| "grad_norm": 0.00020184036111459136, | |
| "learning_rate": 1.709804781750384e-05, | |
| "loss": 0.0036, | |
| "step": 45000 | |
| }, | |
| { | |
| "epoch": 1.9960517657381005, | |
| "grad_norm": 0.0013119287323206663, | |
| "learning_rate": 1.673247057103166e-05, | |
| "loss": 0.0084, | |
| "step": 45500 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.7343463107319165, | |
| "eval_f1": 0.7729732777088736, | |
| "eval_loss": 1.7308247089385986, | |
| "eval_precision": 0.8787412677049286, | |
| "eval_recall": 0.6899310622452574, | |
| "eval_runtime": 455.1243, | |
| "eval_samples_per_second": 199.842, | |
| "eval_steps_per_second": 24.982, | |
| "step": 45590 | |
| }, | |
| { | |
| "epoch": 2.017986400526431, | |
| "grad_norm": 0.0002151148219127208, | |
| "learning_rate": 1.636689332455948e-05, | |
| "loss": 0.0033, | |
| "step": 46000 | |
| }, | |
| { | |
| "epoch": 2.039921035314762, | |
| "grad_norm": 0.0002721097262110561, | |
| "learning_rate": 1.60013160780873e-05, | |
| "loss": 0.0027, | |
| "step": 46500 | |
| }, | |
| { | |
| "epoch": 2.0618556701030926, | |
| "grad_norm": 0.00038531862082891166, | |
| "learning_rate": 1.5635738831615122e-05, | |
| "loss": 0.0053, | |
| "step": 47000 | |
| }, | |
| { | |
| "epoch": 2.0837903048914237, | |
| "grad_norm": 0.00017175752145703882, | |
| "learning_rate": 1.5270161585142942e-05, | |
| "loss": 0.001, | |
| "step": 47500 | |
| }, | |
| { | |
| "epoch": 2.1057249396797544, | |
| "grad_norm": 0.00015803653514012694, | |
| "learning_rate": 1.4904584338670763e-05, | |
| "loss": 0.0056, | |
| "step": 48000 | |
| }, | |
| { | |
| "epoch": 2.127659574468085, | |
| "grad_norm": 0.0005274795112200081, | |
| "learning_rate": 1.4539007092198581e-05, | |
| "loss": 0.0066, | |
| "step": 48500 | |
| }, | |
| { | |
| "epoch": 2.1495942092564158, | |
| "grad_norm": 0.0038681207224726677, | |
| "learning_rate": 1.4173429845726402e-05, | |
| "loss": 0.0036, | |
| "step": 49000 | |
| }, | |
| { | |
| "epoch": 2.1715288440447464, | |
| "grad_norm": 0.0002383293176535517, | |
| "learning_rate": 1.3807852599254223e-05, | |
| "loss": 0.0048, | |
| "step": 49500 | |
| }, | |
| { | |
| "epoch": 2.1934634788330776, | |
| "grad_norm": 0.00013641268014907837, | |
| "learning_rate": 1.3442275352782043e-05, | |
| "loss": 0.0039, | |
| "step": 50000 | |
| }, | |
| { | |
| "epoch": 2.2153981136214083, | |
| "grad_norm": 0.0001230493508046493, | |
| "learning_rate": 1.3076698106309864e-05, | |
| "loss": 0.0, | |
| "step": 50500 | |
| }, | |
| { | |
| "epoch": 2.237332748409739, | |
| "grad_norm": 0.00011413154425099492, | |
| "learning_rate": 1.2711120859837682e-05, | |
| "loss": 0.0018, | |
| "step": 51000 | |
| }, | |
| { | |
| "epoch": 2.2592673831980696, | |
| "grad_norm": 0.00016245432198047638, | |
| "learning_rate": 1.2345543613365505e-05, | |
| "loss": 0.008, | |
| "step": 51500 | |
| }, | |
| { | |
| "epoch": 2.2812020179864003, | |
| "grad_norm": 0.00015192371211014688, | |
| "learning_rate": 1.1979966366893325e-05, | |
| "loss": 0.0051, | |
| "step": 52000 | |
| }, | |
| { | |
| "epoch": 2.3031366527747315, | |
| "grad_norm": 0.00011112165520898998, | |
| "learning_rate": 1.1614389120421146e-05, | |
| "loss": 0.0021, | |
| "step": 52500 | |
| }, | |
| { | |
| "epoch": 2.325071287563062, | |
| "grad_norm": 0.00015647824329789728, | |
| "learning_rate": 1.1248811873948966e-05, | |
| "loss": 0.0023, | |
| "step": 53000 | |
| }, | |
| { | |
| "epoch": 2.347005922351393, | |
| "grad_norm": 0.0360456220805645, | |
| "learning_rate": 1.0883234627476787e-05, | |
| "loss": 0.0018, | |
| "step": 53500 | |
| }, | |
| { | |
| "epoch": 2.3689405571397235, | |
| "grad_norm": 9.487931674811989e-05, | |
| "learning_rate": 1.0517657381004607e-05, | |
| "loss": 0.0069, | |
| "step": 54000 | |
| }, | |
| { | |
| "epoch": 2.390875191928054, | |
| "grad_norm": 0.0001035909226629883, | |
| "learning_rate": 1.0152080134532426e-05, | |
| "loss": 0.007, | |
| "step": 54500 | |
| }, | |
| { | |
| "epoch": 2.4128098267163853, | |
| "grad_norm": 0.00010311613004887477, | |
| "learning_rate": 9.786502888060247e-06, | |
| "loss": 0.0017, | |
| "step": 55000 | |
| }, | |
| { | |
| "epoch": 2.434744461504716, | |
| "grad_norm": 0.0001775302371243015, | |
| "learning_rate": 9.420925641588069e-06, | |
| "loss": 0.0031, | |
| "step": 55500 | |
| }, | |
| { | |
| "epoch": 2.4566790962930467, | |
| "grad_norm": 0.00014991634816396981, | |
| "learning_rate": 9.055348395115888e-06, | |
| "loss": 0.0044, | |
| "step": 56000 | |
| }, | |
| { | |
| "epoch": 2.4786137310813774, | |
| "grad_norm": 8.863507537171245e-05, | |
| "learning_rate": 8.68977114864371e-06, | |
| "loss": 0.0027, | |
| "step": 56500 | |
| }, | |
| { | |
| "epoch": 2.500548365869708, | |
| "grad_norm": 8.503735443810001e-05, | |
| "learning_rate": 8.32419390217153e-06, | |
| "loss": 0.002, | |
| "step": 57000 | |
| }, | |
| { | |
| "epoch": 2.522483000658039, | |
| "grad_norm": 0.00013621764082927257, | |
| "learning_rate": 7.95861665569935e-06, | |
| "loss": 0.0028, | |
| "step": 57500 | |
| }, | |
| { | |
| "epoch": 2.54441763544637, | |
| "grad_norm": 0.00012005689495708793, | |
| "learning_rate": 7.593039409227171e-06, | |
| "loss": 0.0017, | |
| "step": 58000 | |
| }, | |
| { | |
| "epoch": 2.5663522702347006, | |
| "grad_norm": 9.417487308382988e-05, | |
| "learning_rate": 7.2274621627549905e-06, | |
| "loss": 0.009, | |
| "step": 58500 | |
| }, | |
| { | |
| "epoch": 2.5882869050230313, | |
| "grad_norm": 9.123358177021146e-05, | |
| "learning_rate": 6.861884916282811e-06, | |
| "loss": 0.0001, | |
| "step": 59000 | |
| }, | |
| { | |
| "epoch": 2.610221539811362, | |
| "grad_norm": 0.00019517861073836684, | |
| "learning_rate": 6.496307669810631e-06, | |
| "loss": 0.0025, | |
| "step": 59500 | |
| }, | |
| { | |
| "epoch": 2.632156174599693, | |
| "grad_norm": 0.00012165991211077198, | |
| "learning_rate": 6.130730423338452e-06, | |
| "loss": 0.003, | |
| "step": 60000 | |
| }, | |
| { | |
| "epoch": 2.6540908093880238, | |
| "grad_norm": 9.820530249271542e-05, | |
| "learning_rate": 5.7651531768662715e-06, | |
| "loss": 0.0, | |
| "step": 60500 | |
| }, | |
| { | |
| "epoch": 2.6760254441763545, | |
| "grad_norm": 0.00013152812607586384, | |
| "learning_rate": 5.399575930394093e-06, | |
| "loss": 0.0001, | |
| "step": 61000 | |
| }, | |
| { | |
| "epoch": 2.697960078964685, | |
| "grad_norm": 7.07305152900517e-05, | |
| "learning_rate": 5.033998683921913e-06, | |
| "loss": 0.0009, | |
| "step": 61500 | |
| }, | |
| { | |
| "epoch": 2.719894713753016, | |
| "grad_norm": 0.00010383051267126575, | |
| "learning_rate": 4.668421437449734e-06, | |
| "loss": 0.0, | |
| "step": 62000 | |
| }, | |
| { | |
| "epoch": 2.741829348541347, | |
| "grad_norm": 8.12528160167858e-05, | |
| "learning_rate": 4.302844190977553e-06, | |
| "loss": 0.0022, | |
| "step": 62500 | |
| }, | |
| { | |
| "epoch": 2.7637639833296777, | |
| "grad_norm": 8.634920231997967e-05, | |
| "learning_rate": 3.9372669445053745e-06, | |
| "loss": 0.0055, | |
| "step": 63000 | |
| }, | |
| { | |
| "epoch": 2.7856986181180083, | |
| "grad_norm": 0.00012579189206007868, | |
| "learning_rate": 3.5716896980331945e-06, | |
| "loss": 0.0039, | |
| "step": 63500 | |
| }, | |
| { | |
| "epoch": 2.807633252906339, | |
| "grad_norm": 6.862137524876744e-05, | |
| "learning_rate": 3.206112451561015e-06, | |
| "loss": 0.0, | |
| "step": 64000 | |
| }, | |
| { | |
| "epoch": 2.8295678876946697, | |
| "grad_norm": 7.466958777513355e-05, | |
| "learning_rate": 2.8405352050888354e-06, | |
| "loss": 0.0009, | |
| "step": 64500 | |
| }, | |
| { | |
| "epoch": 2.851502522483001, | |
| "grad_norm": 6.52004309813492e-05, | |
| "learning_rate": 2.474957958616656e-06, | |
| "loss": 0.0011, | |
| "step": 65000 | |
| }, | |
| { | |
| "epoch": 2.8734371572713315, | |
| "grad_norm": 9.583857899997383e-05, | |
| "learning_rate": 2.109380712144476e-06, | |
| "loss": 0.0005, | |
| "step": 65500 | |
| }, | |
| { | |
| "epoch": 2.895371792059662, | |
| "grad_norm": 7.938170892884955e-05, | |
| "learning_rate": 1.7438034656722967e-06, | |
| "loss": 0.0022, | |
| "step": 66000 | |
| }, | |
| { | |
| "epoch": 2.917306426847993, | |
| "grad_norm": 9.44985804380849e-05, | |
| "learning_rate": 1.3782262192001172e-06, | |
| "loss": 0.001, | |
| "step": 66500 | |
| }, | |
| { | |
| "epoch": 2.9392410616363236, | |
| "grad_norm": 0.0006269481382332742, | |
| "learning_rate": 1.0126489727279376e-06, | |
| "loss": 0.0011, | |
| "step": 67000 | |
| }, | |
| { | |
| "epoch": 2.9611756964246547, | |
| "grad_norm": 7.675900997128338e-05, | |
| "learning_rate": 6.470717262557578e-07, | |
| "loss": 0.0025, | |
| "step": 67500 | |
| }, | |
| { | |
| "epoch": 2.9831103312129854, | |
| "grad_norm": 9.323044650955126e-05, | |
| "learning_rate": 2.814944797835783e-07, | |
| "loss": 0.0, | |
| "step": 68000 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.7589634206678174, | |
| "eval_f1": 0.8126223301053855, | |
| "eval_loss": 2.1171305179595947, | |
| "eval_precision": 0.828476821192053, | |
| "eval_recall": 0.7973632566799175, | |
| "eval_runtime": 462.1041, | |
| "eval_samples_per_second": 196.824, | |
| "eval_steps_per_second": 24.605, | |
| "step": 68385 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "step": 68385, | |
| "total_flos": 4.2393620250444464e+19, | |
| "train_loss": 0.020372442544634804, | |
| "train_runtime": 6553.5638, | |
| "train_samples_per_second": 83.477, | |
| "train_steps_per_second": 10.435 | |
| } | |
| ], | |
| "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": 4.2393620250444464e+19, | |
| "train_batch_size": 8, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |