{ "best_metric": 47.18389718389718, "best_model_checkpoint": "./whisper-base-common_voice0506/checkpoint-100000", "epoch": 2.167645719983526, "global_step": 100000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.01, "learning_rate": 2.4800000000000004e-06, "loss": 1.7405, "step": 500 }, { "epoch": 0.02, "learning_rate": 4.980000000000001e-06, "loss": 1.311, "step": 1000 }, { "epoch": 0.03, "learning_rate": 7.48e-06, "loss": 1.214, "step": 1500 }, { "epoch": 0.04, "learning_rate": 9.980000000000001e-06, "loss": 1.1417, "step": 2000 }, { "epoch": 0.05, "learning_rate": 9.949387755102042e-06, "loss": 1.1107, "step": 2500 }, { "epoch": 0.07, "learning_rate": 9.898367346938776e-06, "loss": 1.0836, "step": 3000 }, { "epoch": 0.08, "learning_rate": 9.84734693877551e-06, "loss": 1.0591, "step": 3500 }, { "epoch": 0.09, "learning_rate": 9.796326530612246e-06, "loss": 1.0454, "step": 4000 }, { "epoch": 0.1, "learning_rate": 9.745306122448981e-06, "loss": 1.0496, "step": 4500 }, { "epoch": 0.11, "learning_rate": 9.694285714285715e-06, "loss": 0.9996, "step": 5000 }, { "epoch": 0.12, "learning_rate": 9.64326530612245e-06, "loss": 0.9862, "step": 5500 }, { "epoch": 0.13, "learning_rate": 9.592244897959183e-06, "loss": 0.9883, "step": 6000 }, { "epoch": 0.14, "learning_rate": 9.541326530612245e-06, "loss": 0.9572, "step": 6500 }, { "epoch": 0.15, "learning_rate": 9.49030612244898e-06, "loss": 0.9448, "step": 7000 }, { "epoch": 0.16, "learning_rate": 9.439387755102042e-06, "loss": 0.9831, "step": 7500 }, { "epoch": 0.17, "learning_rate": 9.388367346938776e-06, "loss": 0.9387, "step": 8000 }, { "epoch": 0.18, "learning_rate": 9.337346938775511e-06, "loss": 0.9088, "step": 8500 }, { "epoch": 0.2, "learning_rate": 9.286326530612245e-06, "loss": 0.9077, "step": 9000 }, { "epoch": 0.21, "learning_rate": 9.235306122448981e-06, "loss": 0.899, "step": 9500 }, { "epoch": 0.22, "learning_rate": 9.184285714285715e-06, "loss": 0.9046, "step": 10000 }, { "epoch": 0.23, "learning_rate": 9.13326530612245e-06, "loss": 0.9035, "step": 10500 }, { "epoch": 0.24, "learning_rate": 9.082346938775511e-06, "loss": 0.8871, "step": 11000 }, { "epoch": 0.25, "learning_rate": 9.031326530612245e-06, "loss": 0.8914, "step": 11500 }, { "epoch": 0.26, "learning_rate": 8.980306122448981e-06, "loss": 0.8953, "step": 12000 }, { "epoch": 0.27, "learning_rate": 8.929285714285715e-06, "loss": 0.8898, "step": 12500 }, { "epoch": 0.28, "learning_rate": 8.87826530612245e-06, "loss": 0.8539, "step": 13000 }, { "epoch": 0.29, "learning_rate": 8.827244897959185e-06, "loss": 0.8566, "step": 13500 }, { "epoch": 0.3, "learning_rate": 8.776224489795919e-06, "loss": 0.861, "step": 14000 }, { "epoch": 0.31, "learning_rate": 8.725204081632653e-06, "loss": 0.8574, "step": 14500 }, { "epoch": 0.33, "learning_rate": 8.674285714285715e-06, "loss": 0.8854, "step": 15000 }, { "epoch": 0.34, "learning_rate": 8.623265306122449e-06, "loss": 0.8428, "step": 15500 }, { "epoch": 0.35, "learning_rate": 8.572244897959185e-06, "loss": 0.8204, "step": 16000 }, { "epoch": 0.36, "learning_rate": 8.521224489795919e-06, "loss": 0.8466, "step": 16500 }, { "epoch": 0.37, "learning_rate": 8.470204081632653e-06, "loss": 0.8485, "step": 17000 }, { "epoch": 0.38, "learning_rate": 8.419285714285715e-06, "loss": 0.8218, "step": 17500 }, { "epoch": 0.39, "learning_rate": 8.368265306122449e-06, "loss": 0.8029, "step": 18000 }, { "epoch": 0.4, "learning_rate": 8.317244897959185e-06, "loss": 0.8165, "step": 18500 }, { "epoch": 0.41, "learning_rate": 8.266224489795919e-06, "loss": 0.8244, "step": 19000 }, { "epoch": 0.42, "learning_rate": 8.21530612244898e-06, "loss": 0.809, "step": 19500 }, { "epoch": 0.43, "learning_rate": 8.164285714285715e-06, "loss": 0.7981, "step": 20000 }, { "epoch": 0.43, "eval_loss": 0.29890957474708557, "eval_runtime": 2355.2756, "eval_samples_per_second": 4.492, "eval_steps_per_second": 0.141, "eval_wer": 70.46872046872046, "step": 20000 }, { "epoch": 0.44, "learning_rate": 8.11326530612245e-06, "loss": 0.8127, "step": 20500 }, { "epoch": 0.46, "learning_rate": 8.062244897959185e-06, "loss": 0.8024, "step": 21000 }, { "epoch": 0.47, "learning_rate": 8.011224489795919e-06, "loss": 0.8086, "step": 21500 }, { "epoch": 0.48, "learning_rate": 7.960408163265307e-06, "loss": 0.8068, "step": 22000 }, { "epoch": 0.49, "learning_rate": 7.909387755102041e-06, "loss": 0.8069, "step": 22500 }, { "epoch": 0.5, "learning_rate": 7.858367346938777e-06, "loss": 0.7739, "step": 23000 }, { "epoch": 0.51, "learning_rate": 7.80734693877551e-06, "loss": 0.7701, "step": 23500 }, { "epoch": 0.52, "learning_rate": 7.756326530612246e-06, "loss": 0.7903, "step": 24000 }, { "epoch": 0.53, "learning_rate": 7.70530612244898e-06, "loss": 0.7981, "step": 24500 }, { "epoch": 0.54, "learning_rate": 7.654285714285715e-06, "loss": 0.7817, "step": 25000 }, { "epoch": 0.55, "learning_rate": 7.6032653061224495e-06, "loss": 0.8016, "step": 25500 }, { "epoch": 0.56, "learning_rate": 7.552244897959184e-06, "loss": 0.7584, "step": 26000 }, { "epoch": 0.57, "learning_rate": 7.501224489795918e-06, "loss": 0.7484, "step": 26500 }, { "epoch": 0.59, "learning_rate": 7.4503061224489804e-06, "loss": 0.7628, "step": 27000 }, { "epoch": 0.6, "learning_rate": 7.3992857142857145e-06, "loss": 0.7695, "step": 27500 }, { "epoch": 0.61, "learning_rate": 7.348265306122449e-06, "loss": 0.7486, "step": 28000 }, { "epoch": 0.62, "learning_rate": 7.297244897959185e-06, "loss": 0.7492, "step": 28500 }, { "epoch": 0.63, "learning_rate": 7.2463265306122454e-06, "loss": 0.7587, "step": 29000 }, { "epoch": 0.64, "learning_rate": 7.19530612244898e-06, "loss": 0.7381, "step": 29500 }, { "epoch": 0.65, "learning_rate": 7.1443877551020415e-06, "loss": 0.7432, "step": 30000 }, { "epoch": 0.66, "learning_rate": 7.093367346938776e-06, "loss": 0.7431, "step": 30500 }, { "epoch": 0.67, "learning_rate": 7.0423469387755105e-06, "loss": 0.7438, "step": 31000 }, { "epoch": 0.68, "learning_rate": 6.991326530612245e-06, "loss": 0.7516, "step": 31500 }, { "epoch": 0.69, "learning_rate": 6.94030612244898e-06, "loss": 0.7239, "step": 32000 }, { "epoch": 0.7, "learning_rate": 6.889285714285715e-06, "loss": 0.7374, "step": 32500 }, { "epoch": 0.72, "learning_rate": 6.838265306122449e-06, "loss": 0.7216, "step": 33000 }, { "epoch": 0.73, "learning_rate": 6.787244897959184e-06, "loss": 0.7202, "step": 33500 }, { "epoch": 0.74, "learning_rate": 6.736326530612245e-06, "loss": 0.7324, "step": 34000 }, { "epoch": 0.75, "learning_rate": 6.68530612244898e-06, "loss": 0.72, "step": 34500 }, { "epoch": 0.76, "learning_rate": 6.634285714285715e-06, "loss": 0.693, "step": 35000 }, { "epoch": 0.77, "learning_rate": 6.583265306122449e-06, "loss": 0.7279, "step": 35500 }, { "epoch": 0.78, "learning_rate": 6.532244897959185e-06, "loss": 0.7263, "step": 36000 }, { "epoch": 0.79, "learning_rate": 6.481326530612245e-06, "loss": 0.7207, "step": 36500 }, { "epoch": 0.8, "learning_rate": 6.43030612244898e-06, "loss": 0.7084, "step": 37000 }, { "epoch": 0.81, "learning_rate": 6.379285714285715e-06, "loss": 0.6915, "step": 37500 }, { "epoch": 0.82, "learning_rate": 6.32826530612245e-06, "loss": 0.6871, "step": 38000 }, { "epoch": 0.83, "learning_rate": 6.27734693877551e-06, "loss": 0.7237, "step": 38500 }, { "epoch": 0.85, "learning_rate": 6.226326530612246e-06, "loss": 0.6986, "step": 39000 }, { "epoch": 0.86, "learning_rate": 6.17530612244898e-06, "loss": 0.6986, "step": 39500 }, { "epoch": 0.87, "learning_rate": 6.124285714285715e-06, "loss": 0.7128, "step": 40000 }, { "epoch": 0.87, "eval_loss": 0.23399513959884644, "eval_runtime": 2395.947, "eval_samples_per_second": 4.416, "eval_steps_per_second": 0.138, "eval_wer": 60.59346059346059, "step": 40000 }, { "epoch": 0.88, "learning_rate": 6.07326530612245e-06, "loss": 0.6807, "step": 40500 }, { "epoch": 0.89, "learning_rate": 6.022346938775511e-06, "loss": 0.7168, "step": 41000 }, { "epoch": 0.9, "learning_rate": 5.971326530612246e-06, "loss": 0.6922, "step": 41500 }, { "epoch": 0.91, "learning_rate": 5.92030612244898e-06, "loss": 0.7043, "step": 42000 }, { "epoch": 0.92, "learning_rate": 5.869285714285715e-06, "loss": 0.6883, "step": 42500 }, { "epoch": 0.93, "learning_rate": 5.818367346938776e-06, "loss": 0.6933, "step": 43000 }, { "epoch": 0.94, "learning_rate": 5.767346938775511e-06, "loss": 0.6826, "step": 43500 }, { "epoch": 0.95, "learning_rate": 5.716326530612245e-06, "loss": 0.6787, "step": 44000 }, { "epoch": 0.96, "learning_rate": 5.665306122448981e-06, "loss": 0.6764, "step": 44500 }, { "epoch": 0.98, "learning_rate": 5.614387755102041e-06, "loss": 0.6591, "step": 45000 }, { "epoch": 0.99, "learning_rate": 5.563367346938776e-06, "loss": 0.6704, "step": 45500 }, { "epoch": 1.0, "learning_rate": 5.512448979591837e-06, "loss": 0.6814, "step": 46000 }, { "epoch": 1.01, "learning_rate": 5.461428571428572e-06, "loss": 0.6584, "step": 46500 }, { "epoch": 1.02, "learning_rate": 5.410408163265306e-06, "loss": 0.6247, "step": 47000 }, { "epoch": 1.03, "learning_rate": 5.359387755102041e-06, "loss": 0.6341, "step": 47500 }, { "epoch": 1.04, "learning_rate": 5.3083673469387766e-06, "loss": 0.6135, "step": 48000 }, { "epoch": 1.05, "learning_rate": 5.257346938775511e-06, "loss": 0.5983, "step": 48500 }, { "epoch": 1.06, "learning_rate": 5.2063265306122455e-06, "loss": 0.614, "step": 49000 }, { "epoch": 1.07, "learning_rate": 5.1553061224489795e-06, "loss": 0.6115, "step": 49500 }, { "epoch": 1.08, "learning_rate": 5.1043877551020416e-06, "loss": 0.6035, "step": 50000 }, { "epoch": 1.09, "learning_rate": 5.053469387755102e-06, "loss": 0.613, "step": 50500 }, { "epoch": 1.11, "learning_rate": 5.002448979591838e-06, "loss": 0.6306, "step": 51000 }, { "epoch": 1.12, "learning_rate": 4.951428571428572e-06, "loss": 0.6206, "step": 51500 }, { "epoch": 1.13, "learning_rate": 4.900408163265307e-06, "loss": 0.6248, "step": 52000 }, { "epoch": 1.14, "learning_rate": 4.8493877551020415e-06, "loss": 0.599, "step": 52500 }, { "epoch": 1.15, "learning_rate": 4.7983673469387755e-06, "loss": 0.6126, "step": 53000 }, { "epoch": 1.16, "learning_rate": 4.74734693877551e-06, "loss": 0.6153, "step": 53500 }, { "epoch": 1.17, "learning_rate": 4.696326530612245e-06, "loss": 0.6122, "step": 54000 }, { "epoch": 1.18, "learning_rate": 4.64530612244898e-06, "loss": 0.5982, "step": 54500 }, { "epoch": 1.19, "learning_rate": 4.594387755102041e-06, "loss": 0.608, "step": 55000 }, { "epoch": 1.2, "learning_rate": 4.543367346938775e-06, "loss": 0.5832, "step": 55500 }, { "epoch": 1.21, "learning_rate": 4.49234693877551e-06, "loss": 0.6065, "step": 56000 }, { "epoch": 1.22, "learning_rate": 4.441326530612245e-06, "loss": 0.5893, "step": 56500 }, { "epoch": 1.24, "learning_rate": 4.39030612244898e-06, "loss": 0.6243, "step": 57000 }, { "epoch": 1.25, "learning_rate": 4.339285714285715e-06, "loss": 0.6104, "step": 57500 }, { "epoch": 1.26, "learning_rate": 4.288367346938776e-06, "loss": 0.6077, "step": 58000 }, { "epoch": 1.27, "learning_rate": 4.23734693877551e-06, "loss": 0.6018, "step": 58500 }, { "epoch": 1.28, "learning_rate": 4.186326530612245e-06, "loss": 0.592, "step": 59000 }, { "epoch": 1.29, "learning_rate": 4.13530612244898e-06, "loss": 0.6144, "step": 59500 }, { "epoch": 1.3, "learning_rate": 4.084387755102041e-06, "loss": 0.5991, "step": 60000 }, { "epoch": 1.3, "eval_loss": 0.19779521226882935, "eval_runtime": 2343.2046, "eval_samples_per_second": 4.516, "eval_steps_per_second": 0.141, "eval_wer": 53.449253449253455, "step": 60000 }, { "epoch": 1.31, "learning_rate": 4.033367346938776e-06, "loss": 0.6093, "step": 60500 }, { "epoch": 1.32, "learning_rate": 3.98234693877551e-06, "loss": 0.5711, "step": 61000 }, { "epoch": 1.33, "learning_rate": 3.931326530612245e-06, "loss": 0.5823, "step": 61500 }, { "epoch": 1.34, "learning_rate": 3.88030612244898e-06, "loss": 0.5897, "step": 62000 }, { "epoch": 1.35, "learning_rate": 3.829387755102041e-06, "loss": 0.583, "step": 62500 }, { "epoch": 1.37, "learning_rate": 3.778367346938776e-06, "loss": 0.5861, "step": 63000 }, { "epoch": 1.38, "learning_rate": 3.727346938775511e-06, "loss": 0.59, "step": 63500 }, { "epoch": 1.39, "learning_rate": 3.6763265306122453e-06, "loss": 0.5843, "step": 64000 }, { "epoch": 1.4, "learning_rate": 3.6254081632653065e-06, "loss": 0.5879, "step": 64500 }, { "epoch": 1.41, "learning_rate": 3.574387755102041e-06, "loss": 0.5689, "step": 65000 }, { "epoch": 1.42, "learning_rate": 3.5234693877551026e-06, "loss": 0.5889, "step": 65500 }, { "epoch": 1.43, "learning_rate": 3.472448979591837e-06, "loss": 0.5686, "step": 66000 }, { "epoch": 1.44, "learning_rate": 3.4214285714285715e-06, "loss": 0.5652, "step": 66500 }, { "epoch": 1.45, "learning_rate": 3.3704081632653064e-06, "loss": 0.5864, "step": 67000 }, { "epoch": 1.46, "learning_rate": 3.319387755102041e-06, "loss": 0.5742, "step": 67500 }, { "epoch": 1.47, "learning_rate": 3.2683673469387758e-06, "loss": 0.5784, "step": 68000 }, { "epoch": 1.48, "learning_rate": 3.2173469387755107e-06, "loss": 0.574, "step": 68500 }, { "epoch": 1.5, "learning_rate": 3.1663265306122455e-06, "loss": 0.5534, "step": 69000 }, { "epoch": 1.51, "learning_rate": 3.11530612244898e-06, "loss": 0.562, "step": 69500 }, { "epoch": 1.52, "learning_rate": 3.064387755102041e-06, "loss": 0.5634, "step": 70000 }, { "epoch": 1.53, "learning_rate": 3.0133673469387757e-06, "loss": 0.5827, "step": 70500 }, { "epoch": 1.54, "learning_rate": 2.9623469387755106e-06, "loss": 0.5567, "step": 71000 }, { "epoch": 1.55, "learning_rate": 2.911326530612245e-06, "loss": 0.5669, "step": 71500 }, { "epoch": 1.56, "learning_rate": 2.8604081632653062e-06, "loss": 0.5749, "step": 72000 }, { "epoch": 1.57, "learning_rate": 2.809387755102041e-06, "loss": 0.5563, "step": 72500 }, { "epoch": 1.58, "learning_rate": 2.7583673469387756e-06, "loss": 0.562, "step": 73000 }, { "epoch": 1.59, "learning_rate": 2.7073469387755105e-06, "loss": 0.5891, "step": 73500 }, { "epoch": 1.6, "learning_rate": 2.656326530612245e-06, "loss": 0.5775, "step": 74000 }, { "epoch": 1.61, "learning_rate": 2.6053061224489802e-06, "loss": 0.5547, "step": 74500 }, { "epoch": 1.63, "learning_rate": 2.5542857142857147e-06, "loss": 0.5451, "step": 75000 }, { "epoch": 1.64, "learning_rate": 2.503265306122449e-06, "loss": 0.5549, "step": 75500 }, { "epoch": 1.65, "learning_rate": 2.4523469387755104e-06, "loss": 0.5677, "step": 76000 }, { "epoch": 1.66, "learning_rate": 2.4013265306122452e-06, "loss": 0.5576, "step": 76500 }, { "epoch": 1.67, "learning_rate": 2.3503061224489797e-06, "loss": 0.557, "step": 77000 }, { "epoch": 1.68, "learning_rate": 2.2992857142857146e-06, "loss": 0.551, "step": 77500 }, { "epoch": 1.69, "learning_rate": 2.248265306122449e-06, "loss": 0.5566, "step": 78000 }, { "epoch": 1.7, "learning_rate": 2.1972448979591835e-06, "loss": 0.5595, "step": 78500 }, { "epoch": 1.71, "learning_rate": 2.146326530612245e-06, "loss": 0.5449, "step": 79000 }, { "epoch": 1.72, "learning_rate": 2.09530612244898e-06, "loss": 0.5547, "step": 79500 }, { "epoch": 1.73, "learning_rate": 2.0442857142857145e-06, "loss": 0.5493, "step": 80000 }, { "epoch": 1.73, "eval_loss": 0.17888270318508148, "eval_runtime": 2329.5885, "eval_samples_per_second": 4.542, "eval_steps_per_second": 0.142, "eval_wer": 49.234549234549235, "step": 80000 }, { "epoch": 1.74, "learning_rate": 1.993265306122449e-06, "loss": 0.5445, "step": 80500 }, { "epoch": 1.76, "learning_rate": 1.94234693877551e-06, "loss": 0.5569, "step": 81000 }, { "epoch": 1.77, "learning_rate": 1.891326530612245e-06, "loss": 0.5398, "step": 81500 }, { "epoch": 1.78, "learning_rate": 1.8403061224489795e-06, "loss": 0.5406, "step": 82000 }, { "epoch": 1.79, "learning_rate": 1.7892857142857146e-06, "loss": 0.5466, "step": 82500 }, { "epoch": 1.8, "learning_rate": 1.7383673469387758e-06, "loss": 0.5446, "step": 83000 }, { "epoch": 1.81, "learning_rate": 1.6873469387755103e-06, "loss": 0.549, "step": 83500 }, { "epoch": 1.82, "learning_rate": 1.636326530612245e-06, "loss": 0.557, "step": 84000 }, { "epoch": 1.83, "learning_rate": 1.5853061224489796e-06, "loss": 0.5639, "step": 84500 }, { "epoch": 1.84, "learning_rate": 1.5342857142857145e-06, "loss": 0.5496, "step": 85000 }, { "epoch": 1.85, "learning_rate": 1.4833673469387757e-06, "loss": 0.5588, "step": 85500 }, { "epoch": 1.86, "learning_rate": 1.4323469387755104e-06, "loss": 0.5418, "step": 86000 }, { "epoch": 1.88, "learning_rate": 1.381326530612245e-06, "loss": 0.5398, "step": 86500 }, { "epoch": 1.89, "learning_rate": 1.3303061224489797e-06, "loss": 0.55, "step": 87000 }, { "epoch": 1.9, "learning_rate": 1.2792857142857142e-06, "loss": 0.5387, "step": 87500 }, { "epoch": 1.91, "learning_rate": 1.228265306122449e-06, "loss": 0.5475, "step": 88000 }, { "epoch": 1.92, "learning_rate": 1.1772448979591838e-06, "loss": 0.551, "step": 88500 }, { "epoch": 1.93, "learning_rate": 1.1262244897959184e-06, "loss": 0.5529, "step": 89000 }, { "epoch": 1.94, "learning_rate": 1.0752040816326531e-06, "loss": 0.5414, "step": 89500 }, { "epoch": 1.95, "learning_rate": 1.0241836734693878e-06, "loss": 0.5238, "step": 90000 }, { "epoch": 1.96, "learning_rate": 9.732653061224492e-07, "loss": 0.547, "step": 90500 }, { "epoch": 1.97, "learning_rate": 9.223469387755103e-07, "loss": 0.546, "step": 91000 }, { "epoch": 1.98, "learning_rate": 8.71326530612245e-07, "loss": 0.536, "step": 91500 }, { "epoch": 1.99, "learning_rate": 8.203061224489797e-07, "loss": 0.5378, "step": 92000 }, { "epoch": 2.01, "learning_rate": 7.692857142857143e-07, "loss": 0.5174, "step": 92500 }, { "epoch": 2.02, "learning_rate": 7.18265306122449e-07, "loss": 0.5112, "step": 93000 }, { "epoch": 2.03, "learning_rate": 6.672448979591838e-07, "loss": 0.5022, "step": 93500 }, { "epoch": 2.04, "learning_rate": 6.162244897959183e-07, "loss": 0.5009, "step": 94000 }, { "epoch": 2.05, "learning_rate": 5.652040816326531e-07, "loss": 0.5296, "step": 94500 }, { "epoch": 2.06, "learning_rate": 5.142857142857143e-07, "loss": 0.5133, "step": 95000 }, { "epoch": 2.07, "learning_rate": 4.63265306122449e-07, "loss": 0.5038, "step": 95500 }, { "epoch": 2.08, "learning_rate": 4.1224489795918373e-07, "loss": 0.5032, "step": 96000 }, { "epoch": 2.09, "learning_rate": 3.612244897959184e-07, "loss": 0.4894, "step": 96500 }, { "epoch": 2.1, "learning_rate": 3.102040816326531e-07, "loss": 0.4838, "step": 97000 }, { "epoch": 2.11, "learning_rate": 2.5918367346938776e-07, "loss": 0.496, "step": 97500 }, { "epoch": 2.12, "learning_rate": 2.0826530612244898e-07, "loss": 0.486, "step": 98000 }, { "epoch": 2.14, "learning_rate": 1.5724489795918368e-07, "loss": 0.4996, "step": 98500 }, { "epoch": 2.15, "learning_rate": 1.0622448979591837e-07, "loss": 0.5156, "step": 99000 }, { "epoch": 2.16, "learning_rate": 5.520408163265306e-08, "loss": 0.5089, "step": 99500 }, { "epoch": 2.17, "learning_rate": 4.285714285714286e-09, "loss": 0.4825, "step": 100000 }, { "epoch": 2.17, "eval_loss": 0.17133809626102448, "eval_runtime": 2407.7066, "eval_samples_per_second": 4.395, "eval_steps_per_second": 0.137, "eval_wer": 47.18389718389718, "step": 100000 } ], "max_steps": 100000, "num_train_epochs": 3, "total_flos": 1.037753452560384e+20, "trial_name": null, "trial_params": null }