{ "best_metric": 1.4961477518081665, "best_model_checkpoint": "ai-light-dance_singing_ft_pretrain_wav2vec2-large-lv60/checkpoint-4416", "epoch": 9.998756078253987, "global_step": 5520, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.02, "learning_rate": 4.999980160665104e-05, "loss": 14.366, "step": 10 }, { "epoch": 0.04, "learning_rate": 4.999882989170569e-05, "loss": 2.8135, "step": 20 }, { "epoch": 0.05, "learning_rate": 4.999704844700443e-05, "loss": 2.7622, "step": 30 }, { "epoch": 0.07, "learning_rate": 4.999445733024951e-05, "loss": 2.5514, "step": 40 }, { "epoch": 0.09, "learning_rate": 4.9991056625369036e-05, "loss": 2.5963, "step": 50 }, { "epoch": 0.11, "learning_rate": 4.998684644251424e-05, "loss": 2.6667, "step": 60 }, { "epoch": 0.13, "learning_rate": 4.998182691805593e-05, "loss": 2.4731, "step": 70 }, { "epoch": 0.14, "learning_rate": 4.9975998214580034e-05, "loss": 2.3348, "step": 80 }, { "epoch": 0.16, "learning_rate": 4.996936052088237e-05, "loss": 2.5527, "step": 90 }, { "epoch": 0.18, "learning_rate": 4.9961914051962536e-05, "loss": 2.4417, "step": 100 }, { "epoch": 0.2, "learning_rate": 4.995365904901691e-05, "loss": 2.3026, "step": 110 }, { "epoch": 0.22, "learning_rate": 4.994459577943088e-05, "loss": 2.2979, "step": 120 }, { "epoch": 0.24, "learning_rate": 4.993472453677014e-05, "loss": 2.2495, "step": 130 }, { "epoch": 0.25, "learning_rate": 4.992404564077122e-05, "loss": 2.3131, "step": 140 }, { "epoch": 0.27, "learning_rate": 4.991255943733109e-05, "loss": 2.2481, "step": 150 }, { "epoch": 0.29, "learning_rate": 4.9900266298496004e-05, "loss": 2.0497, "step": 160 }, { "epoch": 0.31, "learning_rate": 4.98871666224494e-05, "loss": 2.0117, "step": 170 }, { "epoch": 0.33, "learning_rate": 4.987326083349904e-05, "loss": 1.9544, "step": 180 }, { "epoch": 0.34, "learning_rate": 4.9858549382063255e-05, "loss": 2.2149, "step": 190 }, { "epoch": 0.36, "learning_rate": 4.984303274465635e-05, "loss": 1.947, "step": 200 }, { "epoch": 0.38, "learning_rate": 4.982671142387316e-05, "loss": 2.0139, "step": 210 }, { "epoch": 0.4, "learning_rate": 4.980958594837279e-05, "loss": 2.2968, "step": 220 }, { "epoch": 0.42, "learning_rate": 4.97916568728615e-05, "loss": 2.1341, "step": 230 }, { "epoch": 0.43, "learning_rate": 4.9772924778074706e-05, "loss": 1.8847, "step": 240 }, { "epoch": 0.45, "learning_rate": 4.975339027075817e-05, "loss": 1.8361, "step": 250 }, { "epoch": 0.47, "learning_rate": 4.97330539836484e-05, "loss": 1.9505, "step": 260 }, { "epoch": 0.49, "learning_rate": 4.971191657545209e-05, "loss": 1.9075, "step": 270 }, { "epoch": 0.51, "learning_rate": 4.9689978730824825e-05, "loss": 1.9039, "step": 280 }, { "epoch": 0.52, "learning_rate": 4.9667241160348876e-05, "loss": 1.8887, "step": 290 }, { "epoch": 0.54, "learning_rate": 4.964370460051022e-05, "loss": 1.7909, "step": 300 }, { "epoch": 0.56, "learning_rate": 4.961936981367465e-05, "loss": 1.801, "step": 310 }, { "epoch": 0.58, "learning_rate": 4.9594237588063094e-05, "loss": 1.7252, "step": 320 }, { "epoch": 0.6, "learning_rate": 4.9568308737726096e-05, "loss": 1.7617, "step": 330 }, { "epoch": 0.62, "learning_rate": 4.954158410251744e-05, "loss": 1.7578, "step": 340 }, { "epoch": 0.63, "learning_rate": 4.9514064548066904e-05, "loss": 1.652, "step": 350 }, { "epoch": 0.65, "learning_rate": 4.948575096575232e-05, "loss": 1.6462, "step": 360 }, { "epoch": 0.67, "learning_rate": 4.94566442726706e-05, "loss": 1.79, "step": 370 }, { "epoch": 0.69, "learning_rate": 4.9426745411608114e-05, "loss": 1.7842, "step": 380 }, { "epoch": 0.71, "learning_rate": 4.939605535101006e-05, "loss": 1.7447, "step": 390 }, { "epoch": 0.72, "learning_rate": 4.936457508494918e-05, "loss": 1.667, "step": 400 }, { "epoch": 0.74, "learning_rate": 4.933230563309355e-05, "loss": 1.8771, "step": 410 }, { "epoch": 0.76, "learning_rate": 4.929924804067349e-05, "loss": 1.6324, "step": 420 }, { "epoch": 0.78, "learning_rate": 4.926540337844775e-05, "loss": 1.7609, "step": 430 }, { "epoch": 0.8, "learning_rate": 4.9230772742668866e-05, "loss": 1.7842, "step": 440 }, { "epoch": 0.81, "learning_rate": 4.919535725504757e-05, "loss": 1.6591, "step": 450 }, { "epoch": 0.83, "learning_rate": 4.915915806271654e-05, "loss": 1.8631, "step": 460 }, { "epoch": 0.85, "learning_rate": 4.9122176338193174e-05, "loss": 1.8134, "step": 470 }, { "epoch": 0.87, "learning_rate": 4.908441327934164e-05, "loss": 1.6268, "step": 480 }, { "epoch": 0.89, "learning_rate": 4.904587010933409e-05, "loss": 1.9392, "step": 490 }, { "epoch": 0.9, "learning_rate": 4.9006548076611024e-05, "loss": 1.8746, "step": 500 }, { "epoch": 0.92, "learning_rate": 4.8966448454840854e-05, "loss": 1.6652, "step": 510 }, { "epoch": 0.94, "learning_rate": 4.892557254287864e-05, "loss": 1.6807, "step": 520 }, { "epoch": 0.96, "learning_rate": 4.888392166472405e-05, "loss": 1.7016, "step": 530 }, { "epoch": 0.98, "learning_rate": 4.884149716947845e-05, "loss": 1.6741, "step": 540 }, { "epoch": 1.0, "learning_rate": 4.879830043130118e-05, "loss": 1.6096, "step": 550 }, { "epoch": 1.0, "eval_loss": 1.7650316953659058, "eval_runtime": 245.0033, "eval_samples_per_second": 8.02, "eval_steps_per_second": 4.012, "eval_wer": 1.0052985028648882, "step": 552 }, { "epoch": 1.01, "learning_rate": 4.875433284936512e-05, "loss": 1.7353, "step": 560 }, { "epoch": 1.03, "learning_rate": 4.8709595847811294e-05, "loss": 1.6602, "step": 570 }, { "epoch": 1.05, "learning_rate": 4.866409087570278e-05, "loss": 1.6609, "step": 580 }, { "epoch": 1.07, "learning_rate": 4.861781940697776e-05, "loss": 1.7271, "step": 590 }, { "epoch": 1.09, "learning_rate": 4.8570782940401785e-05, "loss": 1.7166, "step": 600 }, { "epoch": 1.1, "learning_rate": 4.8522982999519234e-05, "loss": 1.7366, "step": 610 }, { "epoch": 1.12, "learning_rate": 4.8474421132603945e-05, "loss": 1.5551, "step": 620 }, { "epoch": 1.14, "learning_rate": 4.8425098912609085e-05, "loss": 1.5944, "step": 630 }, { "epoch": 1.16, "learning_rate": 4.83750179371162e-05, "loss": 1.6248, "step": 640 }, { "epoch": 1.18, "learning_rate": 4.832417982828344e-05, "loss": 1.7225, "step": 650 }, { "epoch": 1.2, "learning_rate": 4.8272586232793085e-05, "loss": 1.6396, "step": 660 }, { "epoch": 1.21, "learning_rate": 4.822023882179811e-05, "loss": 1.6435, "step": 670 }, { "epoch": 1.23, "learning_rate": 4.816713929086812e-05, "loss": 1.6247, "step": 680 }, { "epoch": 1.25, "learning_rate": 4.8113289359934456e-05, "loss": 1.6521, "step": 690 }, { "epoch": 1.27, "learning_rate": 4.8058690773234376e-05, "loss": 1.8499, "step": 700 }, { "epoch": 1.29, "learning_rate": 4.8003345299254685e-05, "loss": 1.8329, "step": 710 }, { "epoch": 1.3, "learning_rate": 4.794725473067437e-05, "loss": 1.7598, "step": 720 }, { "epoch": 1.32, "learning_rate": 4.789042088430658e-05, "loss": 1.5682, "step": 730 }, { "epoch": 1.34, "learning_rate": 4.78328456010397e-05, "loss": 1.6314, "step": 740 }, { "epoch": 1.36, "learning_rate": 4.777453074577784e-05, "loss": 1.6098, "step": 750 }, { "epoch": 1.38, "learning_rate": 4.771547820738035e-05, "loss": 1.7038, "step": 760 }, { "epoch": 1.39, "learning_rate": 4.765568989860065e-05, "loss": 1.6077, "step": 770 }, { "epoch": 1.41, "learning_rate": 4.759516775602428e-05, "loss": 1.6007, "step": 780 }, { "epoch": 1.43, "learning_rate": 4.7533913740006206e-05, "loss": 1.5083, "step": 790 }, { "epoch": 1.45, "learning_rate": 4.7471929834607246e-05, "loss": 1.5897, "step": 800 }, { "epoch": 1.47, "learning_rate": 4.740921804752989e-05, "loss": 1.5798, "step": 810 }, { "epoch": 1.48, "learning_rate": 4.7345780410053185e-05, "loss": 1.6451, "step": 820 }, { "epoch": 1.5, "learning_rate": 4.728161897696704e-05, "loss": 1.7461, "step": 830 }, { "epoch": 1.52, "learning_rate": 4.721673582650558e-05, "loss": 1.6566, "step": 840 }, { "epoch": 1.54, "learning_rate": 4.715113306027987e-05, "loss": 1.6935, "step": 850 }, { "epoch": 1.56, "learning_rate": 4.708481280320984e-05, "loss": 1.8284, "step": 860 }, { "epoch": 1.58, "learning_rate": 4.701777720345546e-05, "loss": 1.8826, "step": 870 }, { "epoch": 1.59, "learning_rate": 4.695002843234714e-05, "loss": 1.8929, "step": 880 }, { "epoch": 1.61, "learning_rate": 4.688156868431543e-05, "loss": 1.6442, "step": 890 }, { "epoch": 1.63, "learning_rate": 4.681240017681993e-05, "loss": 1.6897, "step": 900 }, { "epoch": 1.65, "learning_rate": 4.674252515027744e-05, "loss": 1.6115, "step": 910 }, { "epoch": 1.67, "learning_rate": 4.667194586798942e-05, "loss": 1.6265, "step": 920 }, { "epoch": 1.68, "learning_rate": 4.660066461606867e-05, "loss": 1.5716, "step": 930 }, { "epoch": 1.7, "learning_rate": 4.652868370336528e-05, "loss": 1.6873, "step": 940 }, { "epoch": 1.72, "learning_rate": 4.6456005461391854e-05, "loss": 1.7177, "step": 950 }, { "epoch": 1.74, "learning_rate": 4.638263224424798e-05, "loss": 1.6142, "step": 960 }, { "epoch": 1.76, "learning_rate": 4.6308566428543986e-05, "loss": 1.677, "step": 970 }, { "epoch": 1.77, "learning_rate": 4.623381041332395e-05, "loss": 1.5227, "step": 980 }, { "epoch": 1.79, "learning_rate": 4.615836661998799e-05, "loss": 1.6907, "step": 990 }, { "epoch": 1.81, "learning_rate": 4.608223749221387e-05, "loss": 1.7877, "step": 1000 }, { "epoch": 1.83, "learning_rate": 4.6005425495877794e-05, "loss": 1.6388, "step": 1010 }, { "epoch": 1.85, "learning_rate": 4.5927933118974595e-05, "loss": 1.5614, "step": 1020 }, { "epoch": 1.86, "learning_rate": 4.584976287153706e-05, "loss": 1.574, "step": 1030 }, { "epoch": 1.88, "learning_rate": 4.5770917285554734e-05, "loss": 1.6305, "step": 1040 }, { "epoch": 1.9, "learning_rate": 4.569139891489183e-05, "loss": 1.657, "step": 1050 }, { "epoch": 1.92, "learning_rate": 4.561121033520452e-05, "loss": 1.6514, "step": 1060 }, { "epoch": 1.94, "learning_rate": 4.5530354143857556e-05, "loss": 1.5975, "step": 1070 }, { "epoch": 1.96, "learning_rate": 4.544883295984006e-05, "loss": 1.5849, "step": 1080 }, { "epoch": 1.97, "learning_rate": 4.536664942368077e-05, "loss": 1.5664, "step": 1090 }, { "epoch": 1.99, "learning_rate": 4.528380619736247e-05, "loss": 1.6294, "step": 1100 }, { "epoch": 2.0, "eval_loss": 1.673520803451538, "eval_runtime": 245.2788, "eval_samples_per_second": 8.011, "eval_steps_per_second": 4.008, "eval_wer": 0.9591214342924034, "step": 1104 }, { "epoch": 2.01, "learning_rate": 4.520030596423575e-05, "loss": 1.7583, "step": 1110 }, { "epoch": 2.03, "learning_rate": 4.5116151428932165e-05, "loss": 1.6247, "step": 1120 }, { "epoch": 2.05, "learning_rate": 4.503134531727652e-05, "loss": 1.5874, "step": 1130 }, { "epoch": 2.07, "learning_rate": 4.494589037619867e-05, "loss": 1.518, "step": 1140 }, { "epoch": 2.08, "learning_rate": 4.4859789373644526e-05, "loss": 1.5903, "step": 1150 }, { "epoch": 2.1, "learning_rate": 4.477304509848636e-05, "loss": 1.628, "step": 1160 }, { "epoch": 2.12, "learning_rate": 4.468566036043251e-05, "loss": 1.5482, "step": 1170 }, { "epoch": 2.14, "learning_rate": 4.459763798993636e-05, "loss": 1.5684, "step": 1180 }, { "epoch": 2.16, "learning_rate": 4.450898083810465e-05, "loss": 1.5554, "step": 1190 }, { "epoch": 2.17, "learning_rate": 4.4419691776605146e-05, "loss": 1.6015, "step": 1200 }, { "epoch": 2.19, "learning_rate": 4.432977369757362e-05, "loss": 1.6006, "step": 1210 }, { "epoch": 2.21, "learning_rate": 4.4239229513520156e-05, "loss": 1.5688, "step": 1220 }, { "epoch": 2.23, "learning_rate": 4.41480621572348e-05, "loss": 1.6431, "step": 1230 }, { "epoch": 2.25, "learning_rate": 4.405627458169265e-05, "loss": 1.718, "step": 1240 }, { "epoch": 2.26, "learning_rate": 4.396386975995807e-05, "loss": 1.6374, "step": 1250 }, { "epoch": 2.28, "learning_rate": 4.387085068508852e-05, "loss": 1.5539, "step": 1260 }, { "epoch": 2.3, "learning_rate": 4.377722037003751e-05, "loss": 1.5053, "step": 1270 }, { "epoch": 2.32, "learning_rate": 4.3682981847557094e-05, "loss": 1.5718, "step": 1280 }, { "epoch": 2.34, "learning_rate": 4.358813817009955e-05, "loss": 1.5476, "step": 1290 }, { "epoch": 2.35, "learning_rate": 4.3492692409718594e-05, "loss": 1.6163, "step": 1300 }, { "epoch": 2.37, "learning_rate": 4.3396647657969805e-05, "loss": 1.5646, "step": 1310 }, { "epoch": 2.39, "learning_rate": 4.330000702581053e-05, "loss": 1.535, "step": 1320 }, { "epoch": 2.41, "learning_rate": 4.3202773643499094e-05, "loss": 1.49, "step": 1330 }, { "epoch": 2.43, "learning_rate": 4.310495066049342e-05, "loss": 1.5298, "step": 1340 }, { "epoch": 2.45, "learning_rate": 4.300654124534902e-05, "loss": 1.5872, "step": 1350 }, { "epoch": 2.46, "learning_rate": 4.290754858561637e-05, "loss": 1.5961, "step": 1360 }, { "epoch": 2.48, "learning_rate": 4.2807975887737596e-05, "loss": 1.5535, "step": 1370 }, { "epoch": 2.5, "learning_rate": 4.270782637694273e-05, "loss": 1.6587, "step": 1380 }, { "epoch": 2.52, "learning_rate": 4.260710329714516e-05, "loss": 1.679, "step": 1390 }, { "epoch": 2.54, "learning_rate": 4.250580991083657e-05, "loss": 1.561, "step": 1400 }, { "epoch": 2.55, "learning_rate": 4.2403949498981285e-05, "loss": 1.5632, "step": 1410 }, { "epoch": 2.57, "learning_rate": 4.230152536090999e-05, "loss": 1.4978, "step": 1420 }, { "epoch": 2.59, "learning_rate": 4.219854081421285e-05, "loss": 1.4657, "step": 1430 }, { "epoch": 2.61, "learning_rate": 4.209499919463207e-05, "loss": 1.6614, "step": 1440 }, { "epoch": 2.63, "learning_rate": 4.199090385595383e-05, "loss": 1.6041, "step": 1450 }, { "epoch": 2.64, "learning_rate": 4.1886258169899675e-05, "loss": 1.5801, "step": 1460 }, { "epoch": 2.66, "learning_rate": 4.178106552601727e-05, "loss": 1.5238, "step": 1470 }, { "epoch": 2.68, "learning_rate": 4.167532933157066e-05, "loss": 1.5342, "step": 1480 }, { "epoch": 2.7, "learning_rate": 4.1569053011429855e-05, "loss": 1.5576, "step": 1490 }, { "epoch": 2.72, "learning_rate": 4.146224000795992e-05, "loss": 1.5716, "step": 1500 }, { "epoch": 2.73, "learning_rate": 4.1354893780909476e-05, "loss": 1.5511, "step": 1510 }, { "epoch": 2.75, "learning_rate": 4.1247017807298645e-05, "loss": 1.5813, "step": 1520 }, { "epoch": 2.77, "learning_rate": 4.1138615581306386e-05, "loss": 1.6684, "step": 1530 }, { "epoch": 2.79, "learning_rate": 4.102969061415735e-05, "loss": 1.6, "step": 1540 }, { "epoch": 2.81, "learning_rate": 4.092024643400818e-05, "loss": 1.5761, "step": 1550 }, { "epoch": 2.83, "learning_rate": 4.081028658583314e-05, "loss": 1.5007, "step": 1560 }, { "epoch": 2.84, "learning_rate": 4.069981463130939e-05, "loss": 1.5755, "step": 1570 }, { "epoch": 2.86, "learning_rate": 4.058883414870156e-05, "loss": 1.5925, "step": 1580 }, { "epoch": 2.88, "learning_rate": 4.047734873274586e-05, "loss": 1.5747, "step": 1590 }, { "epoch": 2.9, "learning_rate": 4.0365361994533636e-05, "loss": 1.5586, "step": 1600 }, { "epoch": 2.92, "learning_rate": 4.025287756139445e-05, "loss": 1.5401, "step": 1610 }, { "epoch": 2.93, "learning_rate": 4.013989907677852e-05, "loss": 1.4885, "step": 1620 }, { "epoch": 2.95, "learning_rate": 4.002643020013877e-05, "loss": 1.6404, "step": 1630 }, { "epoch": 2.97, "learning_rate": 3.9912474606812245e-05, "loss": 1.5739, "step": 1640 }, { "epoch": 2.99, "learning_rate": 3.9798035987901096e-05, "loss": 1.5509, "step": 1650 }, { "epoch": 3.0, "eval_loss": 1.617006540298462, "eval_runtime": 245.2584, "eval_samples_per_second": 8.012, "eval_steps_per_second": 4.008, "eval_wer": 0.9852134803770562, "step": 1656 }, { "epoch": 3.01, "learning_rate": 3.968311805015301e-05, "loss": 1.6357, "step": 1660 }, { "epoch": 3.03, "learning_rate": 3.9567724515841164e-05, "loss": 1.564, "step": 1670 }, { "epoch": 3.04, "learning_rate": 3.945185912264362e-05, "loss": 1.4766, "step": 1680 }, { "epoch": 3.06, "learning_rate": 3.9335525623522296e-05, "loss": 1.432, "step": 1690 }, { "epoch": 3.08, "learning_rate": 3.9218727786601393e-05, "loss": 1.5134, "step": 1700 }, { "epoch": 3.1, "learning_rate": 3.9101469395045345e-05, "loss": 1.5439, "step": 1710 }, { "epoch": 3.12, "learning_rate": 3.898375424693625e-05, "loss": 1.5595, "step": 1720 }, { "epoch": 3.13, "learning_rate": 3.886558615515093e-05, "loss": 1.4851, "step": 1730 }, { "epoch": 3.15, "learning_rate": 3.8746968947237297e-05, "loss": 1.5089, "step": 1740 }, { "epoch": 3.17, "learning_rate": 3.862790646529051e-05, "loss": 1.5064, "step": 1750 }, { "epoch": 3.19, "learning_rate": 3.850840256582844e-05, "loss": 1.5289, "step": 1760 }, { "epoch": 3.21, "learning_rate": 3.83884611196668e-05, "loss": 1.549, "step": 1770 }, { "epoch": 3.22, "learning_rate": 3.8268086011793737e-05, "loss": 1.4656, "step": 1780 }, { "epoch": 3.24, "learning_rate": 3.814728114124401e-05, "loss": 1.4998, "step": 1790 }, { "epoch": 3.26, "learning_rate": 3.80260504209727e-05, "loss": 1.5859, "step": 1800 }, { "epoch": 3.28, "learning_rate": 3.790439777772845e-05, "loss": 1.5248, "step": 1810 }, { "epoch": 3.3, "learning_rate": 3.77823271519263e-05, "loss": 1.4839, "step": 1820 }, { "epoch": 3.31, "learning_rate": 3.765984249752004e-05, "loss": 1.592, "step": 1830 }, { "epoch": 3.33, "learning_rate": 3.753694778187411e-05, "loss": 1.5277, "step": 1840 }, { "epoch": 3.35, "learning_rate": 3.741364698563517e-05, "loss": 1.6631, "step": 1850 }, { "epoch": 3.37, "learning_rate": 3.728994410260308e-05, "loss": 1.6362, "step": 1860 }, { "epoch": 3.39, "learning_rate": 3.716584313960159e-05, "loss": 1.4694, "step": 1870 }, { "epoch": 3.41, "learning_rate": 3.7041348116348525e-05, "loss": 1.4702, "step": 1880 }, { "epoch": 3.42, "learning_rate": 3.691646306532564e-05, "loss": 1.4676, "step": 1890 }, { "epoch": 3.44, "learning_rate": 3.679119203164793e-05, "loss": 1.4184, "step": 1900 }, { "epoch": 3.46, "learning_rate": 3.666553907293265e-05, "loss": 1.4747, "step": 1910 }, { "epoch": 3.48, "learning_rate": 3.6539508259167863e-05, "loss": 1.5035, "step": 1920 }, { "epoch": 3.5, "learning_rate": 3.6413103672580664e-05, "loss": 1.5803, "step": 1930 }, { "epoch": 3.51, "learning_rate": 3.628632940750486e-05, "loss": 1.4422, "step": 1940 }, { "epoch": 3.53, "learning_rate": 3.615918957024845e-05, "loss": 1.4767, "step": 1950 }, { "epoch": 3.55, "learning_rate": 3.603168827896055e-05, "loss": 1.5398, "step": 1960 }, { "epoch": 3.57, "learning_rate": 3.590382966349804e-05, "loss": 1.5347, "step": 1970 }, { "epoch": 3.59, "learning_rate": 3.577561786529177e-05, "loss": 1.4277, "step": 1980 }, { "epoch": 3.6, "learning_rate": 3.5647057037212436e-05, "loss": 1.4241, "step": 1990 }, { "epoch": 3.62, "learning_rate": 3.551815134343606e-05, "loss": 1.4842, "step": 2000 }, { "epoch": 3.64, "learning_rate": 3.53889049593091e-05, "loss": 1.4583, "step": 2010 }, { "epoch": 3.66, "learning_rate": 3.525932207121323e-05, "loss": 1.4894, "step": 2020 }, { "epoch": 3.68, "learning_rate": 3.51294068764297e-05, "loss": 1.5002, "step": 2030 }, { "epoch": 3.69, "learning_rate": 3.499916358300343e-05, "loss": 1.6458, "step": 2040 }, { "epoch": 3.71, "learning_rate": 3.486859640960668e-05, "loss": 1.5541, "step": 2050 }, { "epoch": 3.73, "learning_rate": 3.4737709585402416e-05, "loss": 1.6797, "step": 2060 }, { "epoch": 3.75, "learning_rate": 3.460650734990731e-05, "loss": 1.5946, "step": 2070 }, { "epoch": 3.77, "learning_rate": 3.4474993952854425e-05, "loss": 1.7373, "step": 2080 }, { "epoch": 3.79, "learning_rate": 3.434317365405559e-05, "loss": 1.6427, "step": 2090 }, { "epoch": 3.8, "learning_rate": 3.421105072326338e-05, "loss": 1.4509, "step": 2100 }, { "epoch": 3.82, "learning_rate": 3.407862944003282e-05, "loss": 1.4324, "step": 2110 }, { "epoch": 3.84, "learning_rate": 3.394591409358281e-05, "loss": 1.4881, "step": 2120 }, { "epoch": 3.86, "learning_rate": 3.381290898265716e-05, "loss": 1.4738, "step": 2130 }, { "epoch": 3.88, "learning_rate": 3.367961841538534e-05, "loss": 1.5011, "step": 2140 }, { "epoch": 3.89, "learning_rate": 3.3546046709142956e-05, "loss": 1.453, "step": 2150 }, { "epoch": 3.91, "learning_rate": 3.341219819041192e-05, "loss": 1.5015, "step": 2160 }, { "epoch": 3.93, "learning_rate": 3.327807719464029e-05, "loss": 1.4722, "step": 2170 }, { "epoch": 3.95, "learning_rate": 3.314368806610183e-05, "loss": 1.5097, "step": 2180 }, { "epoch": 3.97, "learning_rate": 3.3009035157755344e-05, "loss": 1.4377, "step": 2190 }, { "epoch": 3.98, "learning_rate": 3.287412283110361e-05, "loss": 1.5175, "step": 2200 }, { "epoch": 4.0, "eval_loss": 1.6312003135681152, "eval_runtime": 245.9642, "eval_samples_per_second": 7.989, "eval_steps_per_second": 3.997, "eval_wer": 0.9626024274536381, "step": 2208 }, { "epoch": 4.0, "learning_rate": 3.27389554560522e-05, "loss": 1.5521, "step": 2210 }, { "epoch": 4.02, "learning_rate": 3.260353741076783e-05, "loss": 1.515, "step": 2220 }, { "epoch": 4.04, "learning_rate": 3.246787308153664e-05, "loss": 1.504, "step": 2230 }, { "epoch": 4.06, "learning_rate": 3.233196686262206e-05, "loss": 1.5164, "step": 2240 }, { "epoch": 4.08, "learning_rate": 3.219582315612253e-05, "loss": 1.4216, "step": 2250 }, { "epoch": 4.09, "learning_rate": 3.205944637182883e-05, "loss": 1.5813, "step": 2260 }, { "epoch": 4.11, "learning_rate": 3.192284092708132e-05, "loss": 1.4241, "step": 2270 }, { "epoch": 4.13, "learning_rate": 3.178601124662686e-05, "loss": 1.4727, "step": 2280 }, { "epoch": 4.15, "learning_rate": 3.164896176247542e-05, "loss": 1.4977, "step": 2290 }, { "epoch": 4.17, "learning_rate": 3.151169691375658e-05, "loss": 1.4534, "step": 2300 }, { "epoch": 4.18, "learning_rate": 3.137422114657574e-05, "loss": 1.5037, "step": 2310 }, { "epoch": 4.2, "learning_rate": 3.12365389138701e-05, "loss": 1.4533, "step": 2320 }, { "epoch": 4.22, "learning_rate": 3.109865467526441e-05, "loss": 1.4436, "step": 2330 }, { "epoch": 4.24, "learning_rate": 3.0960572896926535e-05, "loss": 1.4711, "step": 2340 }, { "epoch": 4.26, "learning_rate": 3.0822298051422814e-05, "loss": 1.3989, "step": 2350 }, { "epoch": 4.28, "learning_rate": 3.068383461757314e-05, "loss": 1.4147, "step": 2360 }, { "epoch": 4.29, "learning_rate": 3.0545187080305925e-05, "loss": 1.4504, "step": 2370 }, { "epoch": 4.31, "learning_rate": 3.040635993051283e-05, "loss": 1.5228, "step": 2380 }, { "epoch": 4.33, "learning_rate": 3.0267357664903272e-05, "loss": 1.4371, "step": 2390 }, { "epoch": 4.35, "learning_rate": 3.0128184785858825e-05, "loss": 1.4633, "step": 2400 }, { "epoch": 4.37, "learning_rate": 2.998884580128732e-05, "loss": 1.4331, "step": 2410 }, { "epoch": 4.38, "learning_rate": 2.9849345224476882e-05, "loss": 1.4643, "step": 2420 }, { "epoch": 4.4, "learning_rate": 2.9709687573949714e-05, "loss": 1.4814, "step": 2430 }, { "epoch": 4.42, "learning_rate": 2.9569877373315758e-05, "loss": 1.615, "step": 2440 }, { "epoch": 4.44, "learning_rate": 2.9429919151126144e-05, "loss": 1.539, "step": 2450 }, { "epoch": 4.46, "learning_rate": 2.928981744072655e-05, "loss": 1.4518, "step": 2460 }, { "epoch": 4.47, "learning_rate": 2.914957678011031e-05, "loss": 1.4609, "step": 2470 }, { "epoch": 4.49, "learning_rate": 2.9009201711771482e-05, "loss": 1.507, "step": 2480 }, { "epoch": 4.51, "learning_rate": 2.8868696782557663e-05, "loss": 1.4082, "step": 2490 }, { "epoch": 4.53, "learning_rate": 2.872806654352274e-05, "loss": 1.4045, "step": 2500 }, { "epoch": 4.55, "learning_rate": 2.858731554977948e-05, "loss": 1.4227, "step": 2510 }, { "epoch": 4.56, "learning_rate": 2.844644836035199e-05, "loss": 1.4126, "step": 2520 }, { "epoch": 4.58, "learning_rate": 2.8305469538028007e-05, "loss": 1.3939, "step": 2530 }, { "epoch": 4.6, "learning_rate": 2.8164383649211158e-05, "loss": 1.4062, "step": 2540 }, { "epoch": 4.62, "learning_rate": 2.8023195263773032e-05, "loss": 1.4337, "step": 2550 }, { "epoch": 4.64, "learning_rate": 2.788190895490515e-05, "loss": 1.4434, "step": 2560 }, { "epoch": 4.65, "learning_rate": 2.774052929897083e-05, "loss": 1.4993, "step": 2570 }, { "epoch": 4.67, "learning_rate": 2.7599060875356986e-05, "loss": 1.3943, "step": 2580 }, { "epoch": 4.69, "learning_rate": 2.7457508266325778e-05, "loss": 1.4472, "step": 2590 }, { "epoch": 4.71, "learning_rate": 2.7315876056866175e-05, "loss": 1.4575, "step": 2600 }, { "epoch": 4.73, "learning_rate": 2.7174168834545473e-05, "loss": 1.5409, "step": 2610 }, { "epoch": 4.75, "learning_rate": 2.7032391189360694e-05, "loss": 1.4751, "step": 2620 }, { "epoch": 4.76, "learning_rate": 2.6890547713589887e-05, "loss": 1.5141, "step": 2630 }, { "epoch": 4.78, "learning_rate": 2.674864300164342e-05, "loss": 1.4914, "step": 2640 }, { "epoch": 4.8, "learning_rate": 2.660668164991512e-05, "loss": 1.5067, "step": 2650 }, { "epoch": 4.82, "learning_rate": 2.6464668256633453e-05, "loss": 1.4275, "step": 2660 }, { "epoch": 4.84, "learning_rate": 2.6322607421712507e-05, "loss": 1.4309, "step": 2670 }, { "epoch": 4.85, "learning_rate": 2.618050374660307e-05, "loss": 1.4368, "step": 2680 }, { "epoch": 4.87, "learning_rate": 2.6038361834143536e-05, "loss": 1.4353, "step": 2690 }, { "epoch": 4.89, "learning_rate": 2.5896186288410835e-05, "loss": 1.4194, "step": 2700 }, { "epoch": 4.91, "learning_rate": 2.575398171457132e-05, "loss": 1.4261, "step": 2710 }, { "epoch": 4.93, "learning_rate": 2.5611752718731545e-05, "loss": 1.4367, "step": 2720 }, { "epoch": 4.94, "learning_rate": 2.546950390778916e-05, "loss": 1.4003, "step": 2730 }, { "epoch": 4.96, "learning_rate": 2.5327239889283612e-05, "loss": 1.3966, "step": 2740 }, { "epoch": 4.98, "learning_rate": 2.5184965271246934e-05, "loss": 1.4172, "step": 2750 }, { "epoch": 5.0, "learning_rate": 2.504268466205449e-05, "loss": 1.5267, "step": 2760 }, { "epoch": 5.0, "eval_loss": 1.5032153129577637, "eval_runtime": 256.9784, "eval_samples_per_second": 7.647, "eval_steps_per_second": 3.825, "eval_wer": 0.924927607664346, "step": 2760 }, { "epoch": 5.02, "learning_rate": 2.490040267027568e-05, "loss": 1.4702, "step": 2770 }, { "epoch": 5.04, "learning_rate": 2.475812390452473e-05, "loss": 1.4367, "step": 2780 }, { "epoch": 5.05, "learning_rate": 2.4615852973311344e-05, "loss": 1.3964, "step": 2790 }, { "epoch": 5.07, "learning_rate": 2.4473594484891467e-05, "loss": 1.4511, "step": 2800 }, { "epoch": 5.09, "learning_rate": 2.4331353047117994e-05, "loss": 1.4437, "step": 2810 }, { "epoch": 5.11, "learning_rate": 2.4189133267291562e-05, "loss": 1.4537, "step": 2820 }, { "epoch": 5.13, "learning_rate": 2.4046939752011295e-05, "loss": 1.403, "step": 2830 }, { "epoch": 5.14, "learning_rate": 2.3904777107025563e-05, "loss": 1.4306, "step": 2840 }, { "epoch": 5.16, "learning_rate": 2.3762649937082852e-05, "loss": 1.383, "step": 2850 }, { "epoch": 5.18, "learning_rate": 2.3620562845782572e-05, "loss": 1.4501, "step": 2860 }, { "epoch": 5.2, "learning_rate": 2.347852043542596e-05, "loss": 1.4334, "step": 2870 }, { "epoch": 5.22, "learning_rate": 2.3336527306867002e-05, "loss": 1.4666, "step": 2880 }, { "epoch": 5.24, "learning_rate": 2.3194588059363427e-05, "loss": 1.451, "step": 2890 }, { "epoch": 5.25, "learning_rate": 2.3052707290427705e-05, "loss": 1.3777, "step": 2900 }, { "epoch": 5.27, "learning_rate": 2.2910889595678144e-05, "loss": 1.4644, "step": 2910 }, { "epoch": 5.29, "learning_rate": 2.276913956869003e-05, "loss": 1.3945, "step": 2920 }, { "epoch": 5.31, "learning_rate": 2.2627461800846873e-05, "loss": 1.4466, "step": 2930 }, { "epoch": 5.33, "learning_rate": 2.2485860881191606e-05, "loss": 1.3538, "step": 2940 }, { "epoch": 5.34, "learning_rate": 2.2344341396278043e-05, "loss": 1.4176, "step": 2950 }, { "epoch": 5.36, "learning_rate": 2.220290793002224e-05, "loss": 1.5591, "step": 2960 }, { "epoch": 5.38, "learning_rate": 2.2061565063554064e-05, "loss": 1.4049, "step": 2970 }, { "epoch": 5.4, "learning_rate": 2.1920317375068785e-05, "loss": 1.3985, "step": 2980 }, { "epoch": 5.42, "learning_rate": 2.1779169439678786e-05, "loss": 1.3333, "step": 2990 }, { "epoch": 5.43, "learning_rate": 2.1638125829265385e-05, "loss": 1.3756, "step": 3000 }, { "epoch": 5.45, "learning_rate": 2.1497191112330735e-05, "loss": 1.4875, "step": 3010 }, { "epoch": 5.47, "learning_rate": 2.1356369853849868e-05, "loss": 1.4179, "step": 3020 }, { "epoch": 5.49, "learning_rate": 2.1215666615122778e-05, "loss": 1.3813, "step": 3030 }, { "epoch": 5.51, "learning_rate": 2.1075085953626758e-05, "loss": 1.427, "step": 3040 }, { "epoch": 5.52, "learning_rate": 2.0934632422868722e-05, "loss": 1.4622, "step": 3050 }, { "epoch": 5.54, "learning_rate": 2.0794310572237715e-05, "loss": 1.4232, "step": 3060 }, { "epoch": 5.56, "learning_rate": 2.0654124946857605e-05, "loss": 1.4035, "step": 3070 }, { "epoch": 5.58, "learning_rate": 2.0514080087439773e-05, "loss": 1.4082, "step": 3080 }, { "epoch": 5.6, "learning_rate": 2.037418053013615e-05, "loss": 1.4362, "step": 3090 }, { "epoch": 5.62, "learning_rate": 2.0234430806392197e-05, "loss": 1.3959, "step": 3100 }, { "epoch": 5.63, "learning_rate": 2.009483544280016e-05, "loss": 1.4022, "step": 3110 }, { "epoch": 5.65, "learning_rate": 1.9955398960952458e-05, "loss": 1.4149, "step": 3120 }, { "epoch": 5.67, "learning_rate": 1.981612587729523e-05, "loss": 1.4093, "step": 3130 }, { "epoch": 5.69, "learning_rate": 1.9677020702981997e-05, "loss": 1.3723, "step": 3140 }, { "epoch": 5.71, "learning_rate": 1.953808794372761e-05, "loss": 1.3851, "step": 3150 }, { "epoch": 5.72, "learning_rate": 1.939933209966227e-05, "loss": 1.4208, "step": 3160 }, { "epoch": 5.74, "learning_rate": 1.9260757665185757e-05, "loss": 1.3521, "step": 3170 }, { "epoch": 5.76, "learning_rate": 1.9122369128821877e-05, "loss": 1.3647, "step": 3180 }, { "epoch": 5.78, "learning_rate": 1.8984170973073057e-05, "loss": 1.3867, "step": 3190 }, { "epoch": 5.8, "learning_rate": 1.8846167674275176e-05, "loss": 1.3815, "step": 3200 }, { "epoch": 5.81, "learning_rate": 1.870836370245254e-05, "loss": 1.406, "step": 3210 }, { "epoch": 5.83, "learning_rate": 1.8570763521173122e-05, "loss": 1.4264, "step": 3220 }, { "epoch": 5.85, "learning_rate": 1.8433371587403972e-05, "loss": 1.3882, "step": 3230 }, { "epoch": 5.87, "learning_rate": 1.8296192351366865e-05, "loss": 1.4658, "step": 3240 }, { "epoch": 5.89, "learning_rate": 1.8159230256394145e-05, "loss": 1.5036, "step": 3250 }, { "epoch": 5.9, "learning_rate": 1.802248973878479e-05, "loss": 1.3716, "step": 3260 }, { "epoch": 5.92, "learning_rate": 1.7885975227660738e-05, "loss": 1.3846, "step": 3270 }, { "epoch": 5.94, "learning_rate": 1.7749691144823433e-05, "loss": 1.4218, "step": 3280 }, { "epoch": 5.96, "learning_rate": 1.761364190461056e-05, "loss": 1.341, "step": 3290 }, { "epoch": 5.98, "learning_rate": 1.7477831913753094e-05, "loss": 1.3566, "step": 3300 }, { "epoch": 6.0, "learning_rate": 1.7342265571232564e-05, "loss": 1.4055, "step": 3310 }, { "epoch": 6.0, "eval_loss": 1.6107233762741089, "eval_runtime": 255.0576, "eval_samples_per_second": 7.704, "eval_steps_per_second": 3.854, "eval_wer": 0.9438112254328137, "step": 3312 }, { "epoch": 6.01, "learning_rate": 1.720694726813855e-05, "loss": 1.4852, "step": 3320 }, { "epoch": 6.03, "learning_rate": 1.707188138752645e-05, "loss": 1.4157, "step": 3330 }, { "epoch": 6.05, "learning_rate": 1.6937072304275537e-05, "loss": 1.3686, "step": 3340 }, { "epoch": 6.07, "learning_rate": 1.680252438494723e-05, "loss": 1.3748, "step": 3350 }, { "epoch": 6.09, "learning_rate": 1.6668241987643655e-05, "loss": 1.3679, "step": 3360 }, { "epoch": 6.1, "learning_rate": 1.65342294618665e-05, "loss": 1.357, "step": 3370 }, { "epoch": 6.12, "learning_rate": 1.6400491148376122e-05, "loss": 1.3571, "step": 3380 }, { "epoch": 6.14, "learning_rate": 1.6267031379050954e-05, "loss": 1.3508, "step": 3390 }, { "epoch": 6.16, "learning_rate": 1.613385447674718e-05, "loss": 1.3845, "step": 3400 }, { "epoch": 6.18, "learning_rate": 1.6000964755158722e-05, "loss": 1.3863, "step": 3410 }, { "epoch": 6.2, "learning_rate": 1.5868366518677517e-05, "loss": 1.3524, "step": 3420 }, { "epoch": 6.21, "learning_rate": 1.5736064062254094e-05, "loss": 1.373, "step": 3430 }, { "epoch": 6.23, "learning_rate": 1.5604061671258464e-05, "loss": 1.3702, "step": 3440 }, { "epoch": 6.25, "learning_rate": 1.5472363621341286e-05, "loss": 1.3584, "step": 3450 }, { "epoch": 6.27, "learning_rate": 1.5340974178295417e-05, "loss": 1.3749, "step": 3460 }, { "epoch": 6.29, "learning_rate": 1.5209897597917718e-05, "loss": 1.3644, "step": 3470 }, { "epoch": 6.3, "learning_rate": 1.5079138125871195e-05, "loss": 1.3928, "step": 3480 }, { "epoch": 6.32, "learning_rate": 1.4948699997547506e-05, "loss": 1.3659, "step": 3490 }, { "epoch": 6.34, "learning_rate": 1.4818587437929734e-05, "loss": 1.393, "step": 3500 }, { "epoch": 6.36, "learning_rate": 1.468880466145559e-05, "loss": 1.3721, "step": 3510 }, { "epoch": 6.38, "learning_rate": 1.4559355871880869e-05, "loss": 1.394, "step": 3520 }, { "epoch": 6.39, "learning_rate": 1.4430245262143283e-05, "loss": 1.341, "step": 3530 }, { "epoch": 6.41, "learning_rate": 1.4301477014226664e-05, "loss": 1.4027, "step": 3540 }, { "epoch": 6.43, "learning_rate": 1.4173055299025512e-05, "loss": 1.3432, "step": 3550 }, { "epoch": 6.45, "learning_rate": 1.4044984276209883e-05, "loss": 1.3322, "step": 3560 }, { "epoch": 6.47, "learning_rate": 1.3917268094090663e-05, "loss": 1.3663, "step": 3570 }, { "epoch": 6.48, "learning_rate": 1.3789910889485175e-05, "loss": 1.3612, "step": 3580 }, { "epoch": 6.5, "learning_rate": 1.366291678758323e-05, "loss": 1.3675, "step": 3590 }, { "epoch": 6.52, "learning_rate": 1.3536289901813486e-05, "loss": 1.3394, "step": 3600 }, { "epoch": 6.54, "learning_rate": 1.34100343337102e-05, "loss": 1.4076, "step": 3610 }, { "epoch": 6.56, "learning_rate": 1.3284154172780404e-05, "loss": 1.3392, "step": 3620 }, { "epoch": 6.58, "learning_rate": 1.3158653496371395e-05, "loss": 1.3296, "step": 3630 }, { "epoch": 6.59, "learning_rate": 1.303353636953873e-05, "loss": 1.4404, "step": 3640 }, { "epoch": 6.61, "learning_rate": 1.2908806844914517e-05, "loss": 1.3428, "step": 3650 }, { "epoch": 6.63, "learning_rate": 1.2784468962576136e-05, "loss": 1.34, "step": 3660 }, { "epoch": 6.65, "learning_rate": 1.2660526749915414e-05, "loss": 1.3361, "step": 3670 }, { "epoch": 6.67, "learning_rate": 1.2536984221508164e-05, "loss": 1.3727, "step": 3680 }, { "epoch": 6.68, "learning_rate": 1.2413845378984126e-05, "loss": 1.36, "step": 3690 }, { "epoch": 6.7, "learning_rate": 1.2291114210897387e-05, "loss": 1.4022, "step": 3700 }, { "epoch": 6.72, "learning_rate": 1.2168794692597152e-05, "loss": 1.3365, "step": 3710 }, { "epoch": 6.74, "learning_rate": 1.204689078609902e-05, "loss": 1.339, "step": 3720 }, { "epoch": 6.76, "learning_rate": 1.1925406439956605e-05, "loss": 1.3897, "step": 3730 }, { "epoch": 6.77, "learning_rate": 1.1804345589133683e-05, "loss": 1.5034, "step": 3740 }, { "epoch": 6.79, "learning_rate": 1.1683712154876714e-05, "loss": 1.4006, "step": 3750 }, { "epoch": 6.81, "learning_rate": 1.1563510044587844e-05, "loss": 1.3942, "step": 3760 }, { "epoch": 6.83, "learning_rate": 1.1443743151698295e-05, "loss": 1.3792, "step": 3770 }, { "epoch": 6.85, "learning_rate": 1.1324415355542328e-05, "loss": 1.3591, "step": 3780 }, { "epoch": 6.86, "learning_rate": 1.1205530521231538e-05, "loss": 1.3614, "step": 3790 }, { "epoch": 6.88, "learning_rate": 1.108709249952967e-05, "loss": 1.3251, "step": 3800 }, { "epoch": 6.9, "learning_rate": 1.0969105126727903e-05, "loss": 1.2982, "step": 3810 }, { "epoch": 6.92, "learning_rate": 1.0851572224520556e-05, "loss": 1.3748, "step": 3820 }, { "epoch": 6.94, "learning_rate": 1.073449759988135e-05, "loss": 1.3369, "step": 3830 }, { "epoch": 6.96, "learning_rate": 1.0617885044940063e-05, "loss": 1.3336, "step": 3840 }, { "epoch": 6.97, "learning_rate": 1.0501738336859704e-05, "loss": 1.3476, "step": 3850 }, { "epoch": 6.99, "learning_rate": 1.0386061237714181e-05, "loss": 1.3267, "step": 3860 }, { "epoch": 7.0, "eval_loss": 1.538564920425415, "eval_runtime": 257.7989, "eval_samples_per_second": 7.622, "eval_steps_per_second": 3.813, "eval_wer": 0.9378042018359929, "step": 3864 }, { "epoch": 7.01, "learning_rate": 1.0270857494366442e-05, "loss": 1.5207, "step": 3870 }, { "epoch": 7.03, "learning_rate": 1.0156130838347083e-05, "loss": 1.3194, "step": 3880 }, { "epoch": 7.05, "learning_rate": 1.0041884985733524e-05, "loss": 1.3138, "step": 3890 }, { "epoch": 7.07, "learning_rate": 9.92812363702963e-06, "loss": 1.3357, "step": 3900 }, { "epoch": 7.08, "learning_rate": 9.814850477045811e-06, "loss": 1.3374, "step": 3910 }, { "epoch": 7.1, "learning_rate": 9.702069174779733e-06, "loss": 1.3352, "step": 3920 }, { "epoch": 7.12, "learning_rate": 9.589783383297424e-06, "loss": 1.3285, "step": 3930 }, { "epoch": 7.14, "learning_rate": 9.477996739614975e-06, "loss": 1.3959, "step": 3940 }, { "epoch": 7.16, "learning_rate": 9.366712864580726e-06, "loss": 1.3584, "step": 3950 }, { "epoch": 7.17, "learning_rate": 9.255935362757995e-06, "loss": 1.3495, "step": 3960 }, { "epoch": 7.19, "learning_rate": 9.145667822308285e-06, "loss": 1.3259, "step": 3970 }, { "epoch": 7.21, "learning_rate": 9.035913814875116e-06, "loss": 1.3274, "step": 3980 }, { "epoch": 7.23, "learning_rate": 8.926676895468305e-06, "loss": 1.3257, "step": 3990 }, { "epoch": 7.25, "learning_rate": 8.817960602348824e-06, "loss": 1.3115, "step": 4000 }, { "epoch": 7.26, "learning_rate": 8.709768456914187e-06, "loss": 1.3032, "step": 4010 }, { "epoch": 7.28, "learning_rate": 8.602103963584396e-06, "loss": 1.3578, "step": 4020 }, { "epoch": 7.3, "learning_rate": 8.5056599447696e-06, "loss": 1.5353, "step": 4030 }, { "epoch": 7.32, "learning_rate": 8.399007583813598e-06, "loss": 1.3411, "step": 4040 }, { "epoch": 7.34, "learning_rate": 8.29289294072659e-06, "loss": 1.3024, "step": 4050 }, { "epoch": 7.35, "learning_rate": 8.187319452636822e-06, "loss": 1.3113, "step": 4060 }, { "epoch": 7.37, "learning_rate": 8.082290539144136e-06, "loss": 1.4107, "step": 4070 }, { "epoch": 7.39, "learning_rate": 7.977809602209215e-06, "loss": 1.321, "step": 4080 }, { "epoch": 7.41, "learning_rate": 7.873880026043401e-06, "loss": 1.3047, "step": 4090 }, { "epoch": 7.43, "learning_rate": 7.770505176999066e-06, "loss": 1.316, "step": 4100 }, { "epoch": 7.45, "learning_rate": 7.667688403460585e-06, "loss": 1.3609, "step": 4110 }, { "epoch": 7.46, "learning_rate": 7.565433035735858e-06, "loss": 1.3326, "step": 4120 }, { "epoch": 7.48, "learning_rate": 7.463742385948469e-06, "loss": 1.2952, "step": 4130 }, { "epoch": 7.5, "learning_rate": 7.362619747930385e-06, "loss": 1.2864, "step": 4140 }, { "epoch": 7.52, "learning_rate": 7.2620683971152735e-06, "loss": 1.3464, "step": 4150 }, { "epoch": 7.54, "learning_rate": 7.162091590432407e-06, "loss": 1.3382, "step": 4160 }, { "epoch": 7.55, "learning_rate": 7.062692566201151e-06, "loss": 1.3399, "step": 4170 }, { "epoch": 7.57, "learning_rate": 6.9638745440261084e-06, "loss": 1.3286, "step": 4180 }, { "epoch": 7.59, "learning_rate": 6.865640724692815e-06, "loss": 1.3425, "step": 4190 }, { "epoch": 7.61, "learning_rate": 6.767994290064047e-06, "loss": 1.3574, "step": 4200 }, { "epoch": 7.63, "learning_rate": 6.67093840297679e-06, "loss": 1.3482, "step": 4210 }, { "epoch": 7.64, "learning_rate": 6.574476207139771e-06, "loss": 1.3497, "step": 4220 }, { "epoch": 7.66, "learning_rate": 6.478610827031647e-06, "loss": 1.319, "step": 4230 }, { "epoch": 7.68, "learning_rate": 6.383345367799784e-06, "loss": 1.3026, "step": 4240 }, { "epoch": 7.7, "learning_rate": 6.288682915159686e-06, "loss": 1.3248, "step": 4250 }, { "epoch": 7.72, "learning_rate": 6.194626535295059e-06, "loss": 1.35, "step": 4260 }, { "epoch": 7.73, "learning_rate": 6.101179274758461e-06, "loss": 1.3128, "step": 4270 }, { "epoch": 7.75, "learning_rate": 6.008344160372667e-06, "loss": 1.3057, "step": 4280 }, { "epoch": 7.77, "learning_rate": 5.916124199132597e-06, "loss": 1.3375, "step": 4290 }, { "epoch": 7.79, "learning_rate": 5.824522378107935e-06, "loss": 1.3053, "step": 4300 }, { "epoch": 7.81, "learning_rate": 5.733541664346342e-06, "loss": 1.3479, "step": 4310 }, { "epoch": 7.83, "learning_rate": 5.643185004777404e-06, "loss": 1.3551, "step": 4320 }, { "epoch": 7.84, "learning_rate": 5.553455326117138e-06, "loss": 1.3793, "step": 4330 }, { "epoch": 7.86, "learning_rate": 5.464355534773216e-06, "loss": 1.3298, "step": 4340 }, { "epoch": 7.88, "learning_rate": 5.3758885167507904e-06, "loss": 1.3677, "step": 4350 }, { "epoch": 7.9, "learning_rate": 5.2880571375590655e-06, "loss": 1.3009, "step": 4360 }, { "epoch": 7.92, "learning_rate": 5.2008642421184475e-06, "loss": 1.326, "step": 4370 }, { "epoch": 7.93, "learning_rate": 5.114312654668396e-06, "loss": 1.3168, "step": 4380 }, { "epoch": 7.95, "learning_rate": 5.02840517867596e-06, "loss": 1.383, "step": 4390 }, { "epoch": 7.97, "learning_rate": 4.943144596744956e-06, "loss": 1.3159, "step": 4400 }, { "epoch": 7.99, "learning_rate": 4.8585336705258615e-06, "loss": 1.312, "step": 4410 }, { "epoch": 8.0, "eval_loss": 1.4961477518081665, "eval_runtime": 253.0297, "eval_samples_per_second": 7.766, "eval_steps_per_second": 3.885, "eval_wer": 0.9206456780235353, "step": 4416 }, { "epoch": 8.01, "learning_rate": 4.7745751406263165e-06, "loss": 1.4127, "step": 4420 }, { "epoch": 8.03, "learning_rate": 4.691271726522409e-06, "loss": 1.2819, "step": 4430 }, { "epoch": 8.04, "learning_rate": 4.608626126470561e-06, "loss": 1.3285, "step": 4440 }, { "epoch": 8.06, "learning_rate": 4.526641017420119e-06, "loss": 1.3204, "step": 4450 }, { "epoch": 8.08, "learning_rate": 4.44531905492667e-06, "loss": 1.2922, "step": 4460 }, { "epoch": 8.1, "learning_rate": 4.364662873066016e-06, "loss": 1.3308, "step": 4470 }, { "epoch": 8.12, "learning_rate": 4.284675084348852e-06, "loss": 1.3136, "step": 4480 }, { "epoch": 8.13, "learning_rate": 4.205358279636143e-06, "loss": 1.2926, "step": 4490 }, { "epoch": 8.15, "learning_rate": 4.126715028055225e-06, "loss": 1.3115, "step": 4500 }, { "epoch": 8.17, "learning_rate": 4.048747876916539e-06, "loss": 1.3122, "step": 4510 }, { "epoch": 8.19, "learning_rate": 3.971459351631191e-06, "loss": 1.281, "step": 4520 }, { "epoch": 8.21, "learning_rate": 3.894851955629092e-06, "loss": 1.3512, "step": 4530 }, { "epoch": 8.22, "learning_rate": 3.818928170277911e-06, "loss": 1.3131, "step": 4540 }, { "epoch": 8.24, "learning_rate": 3.743690454802673e-06, "loss": 1.3419, "step": 4550 }, { "epoch": 8.26, "learning_rate": 3.669141246206126e-06, "loss": 1.2918, "step": 4560 }, { "epoch": 8.28, "learning_rate": 3.5952829591897746e-06, "loss": 1.3001, "step": 4570 }, { "epoch": 8.3, "learning_rate": 3.5221179860757154e-06, "loss": 1.2762, "step": 4580 }, { "epoch": 8.31, "learning_rate": 3.4496486967290853e-06, "loss": 1.2924, "step": 4590 }, { "epoch": 8.33, "learning_rate": 3.3778774384813557e-06, "loss": 1.2832, "step": 4600 }, { "epoch": 8.35, "learning_rate": 3.306806536054269e-06, "loss": 1.296, "step": 4610 }, { "epoch": 8.37, "learning_rate": 3.2364382914845504e-06, "loss": 1.317, "step": 4620 }, { "epoch": 8.39, "learning_rate": 3.166774984049342e-06, "loss": 1.3156, "step": 4630 }, { "epoch": 8.41, "learning_rate": 3.0978188701923704e-06, "loss": 1.3157, "step": 4640 }, { "epoch": 8.42, "learning_rate": 3.029572183450868e-06, "loss": 1.3265, "step": 4650 }, { "epoch": 8.44, "learning_rate": 2.962037134383211e-06, "loss": 1.295, "step": 4660 }, { "epoch": 8.46, "learning_rate": 2.8952159104973398e-06, "loss": 1.302, "step": 4670 }, { "epoch": 8.48, "learning_rate": 2.8291106761798885e-06, "loss": 1.2915, "step": 4680 }, { "epoch": 8.5, "learning_rate": 2.763723572626087e-06, "loss": 1.3203, "step": 4690 }, { "epoch": 8.51, "learning_rate": 2.6990567177703986e-06, "loss": 1.3128, "step": 4700 }, { "epoch": 8.53, "learning_rate": 2.6351122062179268e-06, "loss": 1.2987, "step": 4710 }, { "epoch": 8.55, "learning_rate": 2.5718921091765517e-06, "loss": 1.3198, "step": 4720 }, { "epoch": 8.57, "learning_rate": 2.509398474389882e-06, "loss": 1.3123, "step": 4730 }, { "epoch": 8.59, "learning_rate": 2.4476333260708696e-06, "loss": 1.3314, "step": 4740 }, { "epoch": 8.6, "learning_rate": 2.386598664836298e-06, "loss": 1.3204, "step": 4750 }, { "epoch": 8.62, "learning_rate": 2.3262964676419528e-06, "loss": 1.2667, "step": 4760 }, { "epoch": 8.64, "learning_rate": 2.266728687718586e-06, "loss": 1.3397, "step": 4770 }, { "epoch": 8.66, "learning_rate": 2.2078972545086645e-06, "loss": 1.3021, "step": 4780 }, { "epoch": 8.68, "learning_rate": 2.1498040736038587e-06, "loss": 1.3221, "step": 4790 }, { "epoch": 8.69, "learning_rate": 2.0924510266833285e-06, "loss": 1.3425, "step": 4800 }, { "epoch": 8.71, "learning_rate": 2.035839971452766e-06, "loss": 1.2936, "step": 4810 }, { "epoch": 8.73, "learning_rate": 1.9799727415842323e-06, "loss": 1.459, "step": 4820 }, { "epoch": 8.75, "learning_rate": 1.9248511466567644e-06, "loss": 1.3072, "step": 4830 }, { "epoch": 8.77, "learning_rate": 1.8704769720977483e-06, "loss": 1.2915, "step": 4840 }, { "epoch": 8.79, "learning_rate": 1.816851979125106e-06, "loss": 1.292, "step": 4850 }, { "epoch": 8.8, "learning_rate": 1.7639779046902227e-06, "loss": 1.3023, "step": 4860 }, { "epoch": 8.82, "learning_rate": 1.7118564614217208e-06, "loss": 1.3635, "step": 4870 }, { "epoch": 8.84, "learning_rate": 1.6604893375699594e-06, "loss": 1.3174, "step": 4880 }, { "epoch": 8.86, "learning_rate": 1.6098781969523525e-06, "loss": 1.2623, "step": 4890 }, { "epoch": 8.88, "learning_rate": 1.5600246788994938e-06, "loss": 1.3066, "step": 4900 }, { "epoch": 8.89, "learning_rate": 1.5109303982020402e-06, "loss": 1.2935, "step": 4910 }, { "epoch": 8.91, "learning_rate": 1.4625969450584199e-06, "loss": 1.3168, "step": 4920 }, { "epoch": 8.93, "learning_rate": 1.415025885023316e-06, "loss": 1.2979, "step": 4930 }, { "epoch": 8.95, "learning_rate": 1.3682187589569572e-06, "loss": 1.3023, "step": 4940 }, { "epoch": 8.97, "learning_rate": 1.3221770829752234e-06, "loss": 1.297, "step": 4950 }, { "epoch": 8.98, "learning_rate": 1.276902348400505e-06, "loss": 1.3245, "step": 4960 }, { "epoch": 9.0, "eval_loss": 1.5158185958862305, "eval_runtime": 257.6324, "eval_samples_per_second": 7.627, "eval_steps_per_second": 3.816, "eval_wer": 0.9181504528371635, "step": 4968 }, { "epoch": 9.0, "learning_rate": 1.232396021713436e-06, "loss": 1.3873, "step": 4970 }, { "epoch": 9.02, "learning_rate": 1.1886595445053745e-06, "loss": 1.3066, "step": 4980 }, { "epoch": 9.04, "learning_rate": 1.1456943334317045e-06, "loss": 1.292, "step": 4990 }, { "epoch": 9.06, "learning_rate": 1.1035017801659554e-06, "loss": 1.2582, "step": 5000 }, { "epoch": 9.08, "learning_rate": 1.0620832513547258e-06, "loss": 1.275, "step": 5010 }, { "epoch": 9.09, "learning_rate": 1.0214400885734193e-06, "loss": 1.2783, "step": 5020 }, { "epoch": 9.11, "learning_rate": 9.815736082827847e-07, "loss": 1.2711, "step": 5030 }, { "epoch": 9.13, "learning_rate": 9.424851017862696e-07, "loss": 1.2981, "step": 5040 }, { "epoch": 9.15, "learning_rate": 9.041758351882062e-07, "loss": 1.3007, "step": 5050 }, { "epoch": 9.17, "learning_rate": 8.666470493528006e-07, "loss": 1.3151, "step": 5060 }, { "epoch": 9.18, "learning_rate": 8.298999598639279e-07, "loss": 1.3526, "step": 5070 }, { "epoch": 9.2, "learning_rate": 7.939357569857697e-07, "loss": 1.2856, "step": 5080 }, { "epoch": 9.22, "learning_rate": 7.58755605624259e-07, "loss": 1.2705, "step": 5090 }, { "epoch": 9.24, "learning_rate": 7.243606452893486e-07, "loss": 1.2917, "step": 5100 }, { "epoch": 9.26, "learning_rate": 6.907519900580861e-07, "loss": 1.3771, "step": 5110 }, { "epoch": 9.28, "learning_rate": 6.579307285385556e-07, "loss": 1.3387, "step": 5120 }, { "epoch": 9.29, "learning_rate": 6.258979238346036e-07, "loss": 1.291, "step": 5130 }, { "epoch": 9.31, "learning_rate": 5.946546135113862e-07, "loss": 1.3019, "step": 5140 }, { "epoch": 9.33, "learning_rate": 5.642018095618012e-07, "loss": 1.2926, "step": 5150 }, { "epoch": 9.35, "learning_rate": 5.345404983736701e-07, "loss": 1.281, "step": 5160 }, { "epoch": 9.37, "learning_rate": 5.056716406978135e-07, "loss": 1.3013, "step": 5170 }, { "epoch": 9.38, "learning_rate": 4.775961716169203e-07, "loss": 1.2892, "step": 5180 }, { "epoch": 9.4, "learning_rate": 4.50315000515264e-07, "loss": 1.2909, "step": 5190 }, { "epoch": 9.42, "learning_rate": 4.2382901104924265e-07, "loss": 1.2809, "step": 5200 }, { "epoch": 9.44, "learning_rate": 3.981390611187602e-07, "loss": 1.2743, "step": 5210 }, { "epoch": 9.46, "learning_rate": 3.732459828394402e-07, "loss": 1.2875, "step": 5220 }, { "epoch": 9.47, "learning_rate": 3.491505825156671e-07, "loss": 1.2777, "step": 5230 }, { "epoch": 9.49, "learning_rate": 3.2585364061447634e-07, "loss": 1.3026, "step": 5240 }, { "epoch": 9.51, "learning_rate": 3.0335591174027453e-07, "loss": 1.2666, "step": 5250 }, { "epoch": 9.53, "learning_rate": 2.8165812461038165e-07, "loss": 1.2926, "step": 5260 }, { "epoch": 9.55, "learning_rate": 2.607609820314494e-07, "loss": 1.3371, "step": 5270 }, { "epoch": 9.56, "learning_rate": 2.40665160876688e-07, "loss": 1.482, "step": 5280 }, { "epoch": 9.58, "learning_rate": 2.2137131206393103e-07, "loss": 1.266, "step": 5290 }, { "epoch": 9.6, "learning_rate": 2.0288006053457708e-07, "loss": 1.2813, "step": 5300 }, { "epoch": 9.62, "learning_rate": 1.8519200523331448e-07, "loss": 1.3144, "step": 5310 }, { "epoch": 9.64, "learning_rate": 1.6830771908875054e-07, "loss": 1.304, "step": 5320 }, { "epoch": 9.65, "learning_rate": 1.522277489948376e-07, "loss": 1.3071, "step": 5330 }, { "epoch": 9.67, "learning_rate": 1.3695261579316777e-07, "loss": 1.2871, "step": 5340 }, { "epoch": 9.69, "learning_rate": 1.2248281425608632e-07, "loss": 1.3007, "step": 5350 }, { "epoch": 9.71, "learning_rate": 1.0881881307069075e-07, "loss": 1.2939, "step": 5360 }, { "epoch": 9.73, "learning_rate": 9.596105482362338e-08, "loss": 1.2836, "step": 5370 }, { "epoch": 9.75, "learning_rate": 8.390995598676066e-08, "loss": 1.3011, "step": 5380 }, { "epoch": 9.76, "learning_rate": 7.266590690370445e-08, "loss": 1.304, "step": 5390 }, { "epoch": 9.78, "learning_rate": 6.222927177714222e-08, "loss": 1.3162, "step": 5400 }, { "epoch": 9.8, "learning_rate": 5.260038865706196e-08, "loss": 1.3088, "step": 5410 }, { "epoch": 9.82, "learning_rate": 4.377956942978323e-08, "loss": 1.3107, "step": 5420 }, { "epoch": 9.84, "learning_rate": 3.5767099807873516e-08, "loss": 1.2968, "step": 5430 }, { "epoch": 9.85, "learning_rate": 2.8563239320880673e-08, "loss": 1.282, "step": 5440 }, { "epoch": 9.87, "learning_rate": 2.216822130693408e-08, "loss": 1.3101, "step": 5450 }, { "epoch": 9.89, "learning_rate": 1.658225290518678e-08, "loss": 1.2809, "step": 5460 }, { "epoch": 9.91, "learning_rate": 1.180551504909866e-08, "loss": 1.3, "step": 5470 }, { "epoch": 9.93, "learning_rate": 7.838162460582777e-09, "loss": 1.2829, "step": 5480 }, { "epoch": 9.94, "learning_rate": 4.680323644989937e-09, "loss": 1.2834, "step": 5490 }, { "epoch": 9.96, "learning_rate": 2.33210088694813e-09, "loss": 1.293, "step": 5500 }, { "epoch": 9.98, "learning_rate": 7.935702470568407e-10, "loss": 1.3203, "step": 5510 }, { "epoch": 10.0, "learning_rate": 6.478155940570218e-11, "loss": 1.2885, "step": 5520 }, { "epoch": 10.0, "eval_loss": 1.52957022190094, "eval_runtime": 256.7081, "eval_samples_per_second": 7.655, "eval_steps_per_second": 3.829, "eval_wer": 0.9229868769638346, "step": 5520 }, { "epoch": 10.0, "step": 5520, "total_flos": 5.359318129634805e+19, "train_loss": 1.5176503131355064, "train_runtime": 38880.092, "train_samples_per_second": 4.549, "train_steps_per_second": 0.142 } ], "max_steps": 5520, "num_train_epochs": 10, "total_flos": 5.359318129634805e+19, "trial_name": null, "trial_params": null }