whisper-large-v3-pt / trainer_state.json
zuazo's picture
End of training
6fa54a6 verified
raw
history blame
No virus
103 kB
{
"best_metric": 4.036736437419906,
"best_model_checkpoint": "./checkpoint-1000",
"epoch": 70.54673721340389,
"eval_steps": 1000,
"global_step": 20000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.09,
"learning_rate": 5e-08,
"loss": 0.9499,
"step": 25
},
{
"epoch": 0.18,
"learning_rate": 1e-07,
"loss": 0.9489,
"step": 50
},
{
"epoch": 0.26,
"learning_rate": 1.5e-07,
"loss": 0.8993,
"step": 75
},
{
"epoch": 0.35,
"learning_rate": 2e-07,
"loss": 0.7839,
"step": 100
},
{
"epoch": 0.44,
"learning_rate": 2.5e-07,
"loss": 0.4434,
"step": 125
},
{
"epoch": 0.53,
"learning_rate": 3e-07,
"loss": 0.2292,
"step": 150
},
{
"epoch": 0.62,
"learning_rate": 3.5e-07,
"loss": 0.1955,
"step": 175
},
{
"epoch": 0.71,
"learning_rate": 4e-07,
"loss": 0.1655,
"step": 200
},
{
"epoch": 0.79,
"learning_rate": 4.5e-07,
"loss": 0.1542,
"step": 225
},
{
"epoch": 0.88,
"learning_rate": 5e-07,
"loss": 0.1466,
"step": 250
},
{
"epoch": 0.97,
"learning_rate": 5.5e-07,
"loss": 0.1382,
"step": 275
},
{
"epoch": 1.06,
"learning_rate": 6e-07,
"loss": 0.1271,
"step": 300
},
{
"epoch": 1.15,
"learning_rate": 6.5e-07,
"loss": 0.1298,
"step": 325
},
{
"epoch": 1.23,
"learning_rate": 7e-07,
"loss": 0.1206,
"step": 350
},
{
"epoch": 1.32,
"learning_rate": 7.5e-07,
"loss": 0.1161,
"step": 375
},
{
"epoch": 1.41,
"learning_rate": 8e-07,
"loss": 0.1346,
"step": 400
},
{
"epoch": 1.5,
"learning_rate": 8.499999999999999e-07,
"loss": 0.1182,
"step": 425
},
{
"epoch": 1.59,
"learning_rate": 9e-07,
"loss": 0.1197,
"step": 450
},
{
"epoch": 1.68,
"learning_rate": 9.499999999999999e-07,
"loss": 0.1226,
"step": 475
},
{
"epoch": 1.76,
"learning_rate": 9.98e-07,
"loss": 0.1183,
"step": 500
},
{
"epoch": 1.85,
"learning_rate": 9.987692307692307e-07,
"loss": 0.1167,
"step": 525
},
{
"epoch": 1.94,
"learning_rate": 9.974871794871794e-07,
"loss": 0.1151,
"step": 550
},
{
"epoch": 2.03,
"learning_rate": 9.962051282051281e-07,
"loss": 0.1115,
"step": 575
},
{
"epoch": 2.12,
"learning_rate": 9.949230769230769e-07,
"loss": 0.0909,
"step": 600
},
{
"epoch": 2.2,
"learning_rate": 9.936410256410256e-07,
"loss": 0.0971,
"step": 625
},
{
"epoch": 2.29,
"learning_rate": 9.923589743589743e-07,
"loss": 0.0973,
"step": 650
},
{
"epoch": 2.38,
"learning_rate": 9.91076923076923e-07,
"loss": 0.0925,
"step": 675
},
{
"epoch": 2.47,
"learning_rate": 9.897948717948718e-07,
"loss": 0.0881,
"step": 700
},
{
"epoch": 2.56,
"learning_rate": 9.885128205128205e-07,
"loss": 0.0982,
"step": 725
},
{
"epoch": 2.65,
"learning_rate": 9.872307692307692e-07,
"loss": 0.0962,
"step": 750
},
{
"epoch": 2.73,
"learning_rate": 9.85948717948718e-07,
"loss": 0.0869,
"step": 775
},
{
"epoch": 2.82,
"learning_rate": 9.846666666666667e-07,
"loss": 0.0908,
"step": 800
},
{
"epoch": 2.91,
"learning_rate": 9.833846153846154e-07,
"loss": 0.0974,
"step": 825
},
{
"epoch": 3.0,
"learning_rate": 9.821025641025641e-07,
"loss": 0.0915,
"step": 850
},
{
"epoch": 3.09,
"learning_rate": 9.808205128205128e-07,
"loss": 0.0807,
"step": 875
},
{
"epoch": 3.17,
"learning_rate": 9.795384615384616e-07,
"loss": 0.0737,
"step": 900
},
{
"epoch": 3.26,
"learning_rate": 9.782564102564103e-07,
"loss": 0.0801,
"step": 925
},
{
"epoch": 3.35,
"learning_rate": 9.76974358974359e-07,
"loss": 0.0704,
"step": 950
},
{
"epoch": 3.44,
"learning_rate": 9.756923076923077e-07,
"loss": 0.0692,
"step": 975
},
{
"epoch": 3.53,
"learning_rate": 9.744102564102563e-07,
"loss": 0.0702,
"step": 1000
},
{
"epoch": 3.53,
"eval_loss": 0.12894539535045624,
"eval_runtime": 2397.6709,
"eval_samples_per_second": 3.784,
"eval_steps_per_second": 0.236,
"eval_wer": 4.036736437419906,
"step": 1000
},
{
"epoch": 3.62,
"learning_rate": 9.73128205128205e-07,
"loss": 0.0732,
"step": 1025
},
{
"epoch": 3.7,
"learning_rate": 9.718461538461537e-07,
"loss": 0.0735,
"step": 1050
},
{
"epoch": 3.79,
"learning_rate": 9.705641025641024e-07,
"loss": 0.0743,
"step": 1075
},
{
"epoch": 3.88,
"learning_rate": 9.692820512820512e-07,
"loss": 0.0768,
"step": 1100
},
{
"epoch": 3.97,
"learning_rate": 9.679999999999999e-07,
"loss": 0.0697,
"step": 1125
},
{
"epoch": 4.06,
"learning_rate": 9.667179487179486e-07,
"loss": 0.0629,
"step": 1150
},
{
"epoch": 4.14,
"learning_rate": 9.654358974358973e-07,
"loss": 0.0555,
"step": 1175
},
{
"epoch": 4.23,
"learning_rate": 9.64153846153846e-07,
"loss": 0.0543,
"step": 1200
},
{
"epoch": 4.32,
"learning_rate": 9.628717948717948e-07,
"loss": 0.0577,
"step": 1225
},
{
"epoch": 4.41,
"learning_rate": 9.615897435897435e-07,
"loss": 0.0596,
"step": 1250
},
{
"epoch": 4.5,
"learning_rate": 9.603076923076922e-07,
"loss": 0.0593,
"step": 1275
},
{
"epoch": 4.59,
"learning_rate": 9.59025641025641e-07,
"loss": 0.058,
"step": 1300
},
{
"epoch": 4.67,
"learning_rate": 9.577435897435897e-07,
"loss": 0.0576,
"step": 1325
},
{
"epoch": 4.76,
"learning_rate": 9.564615384615384e-07,
"loss": 0.0565,
"step": 1350
},
{
"epoch": 4.85,
"learning_rate": 9.551794871794871e-07,
"loss": 0.0574,
"step": 1375
},
{
"epoch": 4.94,
"learning_rate": 9.538974358974359e-07,
"loss": 0.0572,
"step": 1400
},
{
"epoch": 5.03,
"learning_rate": 9.526153846153846e-07,
"loss": 0.0562,
"step": 1425
},
{
"epoch": 5.11,
"learning_rate": 9.513333333333333e-07,
"loss": 0.0443,
"step": 1450
},
{
"epoch": 5.2,
"learning_rate": 9.50051282051282e-07,
"loss": 0.0463,
"step": 1475
},
{
"epoch": 5.29,
"learning_rate": 9.487692307692308e-07,
"loss": 0.043,
"step": 1500
},
{
"epoch": 5.38,
"learning_rate": 9.474871794871795e-07,
"loss": 0.0412,
"step": 1525
},
{
"epoch": 5.47,
"learning_rate": 9.462051282051282e-07,
"loss": 0.0414,
"step": 1550
},
{
"epoch": 5.56,
"learning_rate": 9.449230769230769e-07,
"loss": 0.0493,
"step": 1575
},
{
"epoch": 5.64,
"learning_rate": 9.436410256410255e-07,
"loss": 0.0458,
"step": 1600
},
{
"epoch": 5.73,
"learning_rate": 9.423589743589743e-07,
"loss": 0.0407,
"step": 1625
},
{
"epoch": 5.82,
"learning_rate": 9.41076923076923e-07,
"loss": 0.0447,
"step": 1650
},
{
"epoch": 5.91,
"learning_rate": 9.397948717948717e-07,
"loss": 0.0457,
"step": 1675
},
{
"epoch": 6.0,
"learning_rate": 9.385128205128204e-07,
"loss": 0.045,
"step": 1700
},
{
"epoch": 6.08,
"learning_rate": 9.372307692307692e-07,
"loss": 0.0299,
"step": 1725
},
{
"epoch": 6.17,
"learning_rate": 9.359487179487179e-07,
"loss": 0.0293,
"step": 1750
},
{
"epoch": 6.26,
"learning_rate": 9.346666666666666e-07,
"loss": 0.0303,
"step": 1775
},
{
"epoch": 6.35,
"learning_rate": 9.333846153846153e-07,
"loss": 0.0325,
"step": 1800
},
{
"epoch": 6.44,
"learning_rate": 9.321025641025641e-07,
"loss": 0.0344,
"step": 1825
},
{
"epoch": 6.53,
"learning_rate": 9.308205128205128e-07,
"loss": 0.0304,
"step": 1850
},
{
"epoch": 6.61,
"learning_rate": 9.295384615384615e-07,
"loss": 0.034,
"step": 1875
},
{
"epoch": 6.7,
"learning_rate": 9.282564102564102e-07,
"loss": 0.0341,
"step": 1900
},
{
"epoch": 6.79,
"learning_rate": 9.26974358974359e-07,
"loss": 0.035,
"step": 1925
},
{
"epoch": 6.88,
"learning_rate": 9.256923076923077e-07,
"loss": 0.028,
"step": 1950
},
{
"epoch": 6.97,
"learning_rate": 9.244102564102564e-07,
"loss": 0.0339,
"step": 1975
},
{
"epoch": 7.05,
"learning_rate": 9.231282051282051e-07,
"loss": 0.0247,
"step": 2000
},
{
"epoch": 7.05,
"eval_loss": 0.18059006333351135,
"eval_runtime": 2250.9984,
"eval_samples_per_second": 4.03,
"eval_steps_per_second": 0.252,
"eval_wer": 4.429402293562909,
"step": 2000
},
{
"epoch": 7.14,
"learning_rate": 9.218461538461539e-07,
"loss": 0.0225,
"step": 2025
},
{
"epoch": 7.23,
"learning_rate": 9.205641025641026e-07,
"loss": 0.0218,
"step": 2050
},
{
"epoch": 7.32,
"learning_rate": 9.192820512820513e-07,
"loss": 0.0207,
"step": 2075
},
{
"epoch": 7.41,
"learning_rate": 9.18e-07,
"loss": 0.0206,
"step": 2100
},
{
"epoch": 7.5,
"learning_rate": 9.167179487179488e-07,
"loss": 0.0205,
"step": 2125
},
{
"epoch": 7.58,
"learning_rate": 9.154358974358975e-07,
"loss": 0.0206,
"step": 2150
},
{
"epoch": 7.67,
"learning_rate": 9.141538461538462e-07,
"loss": 0.0205,
"step": 2175
},
{
"epoch": 7.76,
"learning_rate": 9.128717948717947e-07,
"loss": 0.0216,
"step": 2200
},
{
"epoch": 7.85,
"learning_rate": 9.115897435897435e-07,
"loss": 0.0236,
"step": 2225
},
{
"epoch": 7.94,
"learning_rate": 9.103076923076922e-07,
"loss": 0.0229,
"step": 2250
},
{
"epoch": 8.02,
"learning_rate": 9.090256410256409e-07,
"loss": 0.0214,
"step": 2275
},
{
"epoch": 8.11,
"learning_rate": 9.077435897435896e-07,
"loss": 0.0126,
"step": 2300
},
{
"epoch": 8.2,
"learning_rate": 9.064615384615384e-07,
"loss": 0.0143,
"step": 2325
},
{
"epoch": 8.29,
"learning_rate": 9.051794871794871e-07,
"loss": 0.0119,
"step": 2350
},
{
"epoch": 8.38,
"learning_rate": 9.038974358974358e-07,
"loss": 0.0127,
"step": 2375
},
{
"epoch": 8.47,
"learning_rate": 9.026153846153845e-07,
"loss": 0.0146,
"step": 2400
},
{
"epoch": 8.55,
"learning_rate": 9.013333333333333e-07,
"loss": 0.0149,
"step": 2425
},
{
"epoch": 8.64,
"learning_rate": 9.00051282051282e-07,
"loss": 0.0149,
"step": 2450
},
{
"epoch": 8.73,
"learning_rate": 8.987692307692307e-07,
"loss": 0.0136,
"step": 2475
},
{
"epoch": 8.82,
"learning_rate": 8.974871794871794e-07,
"loss": 0.0149,
"step": 2500
},
{
"epoch": 8.91,
"learning_rate": 8.962051282051282e-07,
"loss": 0.0136,
"step": 2525
},
{
"epoch": 8.99,
"learning_rate": 8.949230769230769e-07,
"loss": 0.0149,
"step": 2550
},
{
"epoch": 9.08,
"learning_rate": 8.936410256410256e-07,
"loss": 0.009,
"step": 2575
},
{
"epoch": 9.17,
"learning_rate": 8.924102564102563e-07,
"loss": 0.0074,
"step": 2600
},
{
"epoch": 9.26,
"learning_rate": 8.911282051282051e-07,
"loss": 0.0088,
"step": 2625
},
{
"epoch": 9.35,
"learning_rate": 8.898461538461538e-07,
"loss": 0.0093,
"step": 2650
},
{
"epoch": 9.44,
"learning_rate": 8.885641025641025e-07,
"loss": 0.0087,
"step": 2675
},
{
"epoch": 9.52,
"learning_rate": 8.872820512820512e-07,
"loss": 0.0084,
"step": 2700
},
{
"epoch": 9.61,
"learning_rate": 8.86e-07,
"loss": 0.0093,
"step": 2725
},
{
"epoch": 9.7,
"learning_rate": 8.847179487179487e-07,
"loss": 0.0103,
"step": 2750
},
{
"epoch": 9.79,
"learning_rate": 8.834358974358974e-07,
"loss": 0.0096,
"step": 2775
},
{
"epoch": 9.88,
"learning_rate": 8.821538461538461e-07,
"loss": 0.0097,
"step": 2800
},
{
"epoch": 9.96,
"learning_rate": 8.808717948717949e-07,
"loss": 0.0096,
"step": 2825
},
{
"epoch": 10.05,
"learning_rate": 8.795897435897436e-07,
"loss": 0.009,
"step": 2850
},
{
"epoch": 10.14,
"learning_rate": 8.783076923076923e-07,
"loss": 0.0074,
"step": 2875
},
{
"epoch": 10.23,
"learning_rate": 8.77025641025641e-07,
"loss": 0.0058,
"step": 2900
},
{
"epoch": 10.32,
"learning_rate": 8.757435897435898e-07,
"loss": 0.0065,
"step": 2925
},
{
"epoch": 10.41,
"learning_rate": 8.744615384615385e-07,
"loss": 0.0048,
"step": 2950
},
{
"epoch": 10.49,
"learning_rate": 8.731794871794872e-07,
"loss": 0.0068,
"step": 2975
},
{
"epoch": 10.58,
"learning_rate": 8.718974358974358e-07,
"loss": 0.0074,
"step": 3000
},
{
"epoch": 10.58,
"eval_loss": 0.2821428179740906,
"eval_runtime": 2256.0857,
"eval_samples_per_second": 4.021,
"eval_steps_per_second": 0.251,
"eval_wer": 4.748135247921664,
"step": 3000
},
{
"epoch": 10.67,
"learning_rate": 8.706153846153846e-07,
"loss": 0.0053,
"step": 3025
},
{
"epoch": 10.76,
"learning_rate": 8.693333333333333e-07,
"loss": 0.0053,
"step": 3050
},
{
"epoch": 10.85,
"learning_rate": 8.68051282051282e-07,
"loss": 0.0069,
"step": 3075
},
{
"epoch": 10.93,
"learning_rate": 8.667692307692307e-07,
"loss": 0.0066,
"step": 3100
},
{
"epoch": 11.02,
"learning_rate": 8.654871794871795e-07,
"loss": 0.0053,
"step": 3125
},
{
"epoch": 11.11,
"learning_rate": 8.642051282051282e-07,
"loss": 0.0041,
"step": 3150
},
{
"epoch": 11.2,
"learning_rate": 8.629230769230769e-07,
"loss": 0.0043,
"step": 3175
},
{
"epoch": 11.29,
"learning_rate": 8.616410256410256e-07,
"loss": 0.004,
"step": 3200
},
{
"epoch": 11.38,
"learning_rate": 8.603589743589744e-07,
"loss": 0.0047,
"step": 3225
},
{
"epoch": 11.46,
"learning_rate": 8.590769230769231e-07,
"loss": 0.0044,
"step": 3250
},
{
"epoch": 11.55,
"learning_rate": 8.577948717948718e-07,
"loss": 0.0039,
"step": 3275
},
{
"epoch": 11.64,
"learning_rate": 8.565128205128204e-07,
"loss": 0.0056,
"step": 3300
},
{
"epoch": 11.73,
"learning_rate": 8.552307692307691e-07,
"loss": 0.0052,
"step": 3325
},
{
"epoch": 11.82,
"learning_rate": 8.539487179487179e-07,
"loss": 0.0046,
"step": 3350
},
{
"epoch": 11.9,
"learning_rate": 8.526666666666666e-07,
"loss": 0.0044,
"step": 3375
},
{
"epoch": 11.99,
"learning_rate": 8.513846153846153e-07,
"loss": 0.0057,
"step": 3400
},
{
"epoch": 12.08,
"learning_rate": 8.50102564102564e-07,
"loss": 0.004,
"step": 3425
},
{
"epoch": 12.17,
"learning_rate": 8.488205128205128e-07,
"loss": 0.0031,
"step": 3450
},
{
"epoch": 12.26,
"learning_rate": 8.475384615384615e-07,
"loss": 0.0029,
"step": 3475
},
{
"epoch": 12.35,
"learning_rate": 8.462564102564102e-07,
"loss": 0.0034,
"step": 3500
},
{
"epoch": 12.43,
"learning_rate": 8.45025641025641e-07,
"loss": 0.0035,
"step": 3525
},
{
"epoch": 12.52,
"learning_rate": 8.437435897435897e-07,
"loss": 0.0032,
"step": 3550
},
{
"epoch": 12.61,
"learning_rate": 8.424615384615384e-07,
"loss": 0.0039,
"step": 3575
},
{
"epoch": 12.7,
"learning_rate": 8.411794871794871e-07,
"loss": 0.0038,
"step": 3600
},
{
"epoch": 12.79,
"learning_rate": 8.398974358974359e-07,
"loss": 0.0058,
"step": 3625
},
{
"epoch": 12.87,
"learning_rate": 8.386153846153846e-07,
"loss": 0.0041,
"step": 3650
},
{
"epoch": 12.96,
"learning_rate": 8.373333333333333e-07,
"loss": 0.0028,
"step": 3675
},
{
"epoch": 13.05,
"learning_rate": 8.36051282051282e-07,
"loss": 0.0031,
"step": 3700
},
{
"epoch": 13.14,
"learning_rate": 8.347692307692308e-07,
"loss": 0.002,
"step": 3725
},
{
"epoch": 13.23,
"learning_rate": 8.334871794871795e-07,
"loss": 0.0023,
"step": 3750
},
{
"epoch": 13.32,
"learning_rate": 8.322051282051281e-07,
"loss": 0.0026,
"step": 3775
},
{
"epoch": 13.4,
"learning_rate": 8.309230769230768e-07,
"loss": 0.0029,
"step": 3800
},
{
"epoch": 13.49,
"learning_rate": 8.296410256410256e-07,
"loss": 0.0024,
"step": 3825
},
{
"epoch": 13.58,
"learning_rate": 8.283589743589743e-07,
"loss": 0.0029,
"step": 3850
},
{
"epoch": 13.67,
"learning_rate": 8.27076923076923e-07,
"loss": 0.0023,
"step": 3875
},
{
"epoch": 13.76,
"learning_rate": 8.257948717948717e-07,
"loss": 0.0034,
"step": 3900
},
{
"epoch": 13.84,
"learning_rate": 8.245128205128205e-07,
"loss": 0.0028,
"step": 3925
},
{
"epoch": 13.93,
"learning_rate": 8.232307692307692e-07,
"loss": 0.0031,
"step": 3950
},
{
"epoch": 14.02,
"learning_rate": 8.219487179487179e-07,
"loss": 0.0029,
"step": 3975
},
{
"epoch": 14.11,
"learning_rate": 8.206666666666666e-07,
"loss": 0.0022,
"step": 4000
},
{
"epoch": 14.11,
"eval_loss": 0.3159928321838379,
"eval_runtime": 2240.1095,
"eval_samples_per_second": 4.05,
"eval_steps_per_second": 0.253,
"eval_wer": 4.624913744947919,
"step": 4000
},
{
"epoch": 14.2,
"learning_rate": 8.193846153846154e-07,
"loss": 0.0022,
"step": 4025
},
{
"epoch": 14.29,
"learning_rate": 8.181025641025641e-07,
"loss": 0.0026,
"step": 4050
},
{
"epoch": 14.37,
"learning_rate": 8.168205128205128e-07,
"loss": 0.0035,
"step": 4075
},
{
"epoch": 14.46,
"learning_rate": 8.155384615384615e-07,
"loss": 0.0024,
"step": 4100
},
{
"epoch": 14.55,
"learning_rate": 8.142564102564103e-07,
"loss": 0.0033,
"step": 4125
},
{
"epoch": 14.64,
"learning_rate": 8.12974358974359e-07,
"loss": 0.0028,
"step": 4150
},
{
"epoch": 14.73,
"learning_rate": 8.116923076923077e-07,
"loss": 0.0024,
"step": 4175
},
{
"epoch": 14.81,
"learning_rate": 8.104102564102564e-07,
"loss": 0.0018,
"step": 4200
},
{
"epoch": 14.9,
"learning_rate": 8.091282051282052e-07,
"loss": 0.0026,
"step": 4225
},
{
"epoch": 14.99,
"learning_rate": 8.078461538461539e-07,
"loss": 0.0019,
"step": 4250
},
{
"epoch": 15.08,
"learning_rate": 8.065641025641026e-07,
"loss": 0.0023,
"step": 4275
},
{
"epoch": 15.17,
"learning_rate": 8.052820512820513e-07,
"loss": 0.0019,
"step": 4300
},
{
"epoch": 15.26,
"learning_rate": 8.04e-07,
"loss": 0.0024,
"step": 4325
},
{
"epoch": 15.34,
"learning_rate": 8.027179487179488e-07,
"loss": 0.0033,
"step": 4350
},
{
"epoch": 15.43,
"learning_rate": 8.014358974358974e-07,
"loss": 0.0031,
"step": 4375
},
{
"epoch": 15.52,
"learning_rate": 8.001538461538461e-07,
"loss": 0.0034,
"step": 4400
},
{
"epoch": 15.61,
"learning_rate": 7.988717948717947e-07,
"loss": 0.0037,
"step": 4425
},
{
"epoch": 15.7,
"learning_rate": 7.975897435897435e-07,
"loss": 0.0022,
"step": 4450
},
{
"epoch": 15.78,
"learning_rate": 7.963076923076922e-07,
"loss": 0.0034,
"step": 4475
},
{
"epoch": 15.87,
"learning_rate": 7.950256410256409e-07,
"loss": 0.0028,
"step": 4500
},
{
"epoch": 15.96,
"learning_rate": 7.937435897435896e-07,
"loss": 0.0029,
"step": 4525
},
{
"epoch": 16.05,
"learning_rate": 7.924615384615384e-07,
"loss": 0.0018,
"step": 4550
},
{
"epoch": 16.14,
"learning_rate": 7.911794871794871e-07,
"loss": 0.0016,
"step": 4575
},
{
"epoch": 16.23,
"learning_rate": 7.898974358974358e-07,
"loss": 0.0014,
"step": 4600
},
{
"epoch": 16.31,
"learning_rate": 7.886153846153845e-07,
"loss": 0.0013,
"step": 4625
},
{
"epoch": 16.4,
"learning_rate": 7.873333333333333e-07,
"loss": 0.0013,
"step": 4650
},
{
"epoch": 16.49,
"learning_rate": 7.86051282051282e-07,
"loss": 0.0016,
"step": 4675
},
{
"epoch": 16.58,
"learning_rate": 7.847692307692307e-07,
"loss": 0.0016,
"step": 4700
},
{
"epoch": 16.67,
"learning_rate": 7.834871794871794e-07,
"loss": 0.0013,
"step": 4725
},
{
"epoch": 16.75,
"learning_rate": 7.822051282051282e-07,
"loss": 0.0016,
"step": 4750
},
{
"epoch": 16.84,
"learning_rate": 7.809230769230769e-07,
"loss": 0.0017,
"step": 4775
},
{
"epoch": 16.93,
"learning_rate": 7.796410256410256e-07,
"loss": 0.0015,
"step": 4800
},
{
"epoch": 17.02,
"learning_rate": 7.783589743589743e-07,
"loss": 0.0017,
"step": 4825
},
{
"epoch": 17.11,
"learning_rate": 7.770769230769231e-07,
"loss": 0.0024,
"step": 4850
},
{
"epoch": 17.2,
"learning_rate": 7.757948717948718e-07,
"loss": 0.002,
"step": 4875
},
{
"epoch": 17.28,
"learning_rate": 7.745128205128205e-07,
"loss": 0.0022,
"step": 4900
},
{
"epoch": 17.37,
"learning_rate": 7.732307692307692e-07,
"loss": 0.0013,
"step": 4925
},
{
"epoch": 17.46,
"learning_rate": 7.71948717948718e-07,
"loss": 0.0013,
"step": 4950
},
{
"epoch": 17.55,
"learning_rate": 7.706666666666667e-07,
"loss": 0.0013,
"step": 4975
},
{
"epoch": 17.64,
"learning_rate": 7.693846153846153e-07,
"loss": 0.0016,
"step": 5000
},
{
"epoch": 17.64,
"eval_loss": 0.3260825276374817,
"eval_runtime": 2250.0889,
"eval_samples_per_second": 4.032,
"eval_steps_per_second": 0.252,
"eval_wer": 4.647915092169684,
"step": 5000
},
{
"epoch": 17.72,
"learning_rate": 7.68102564102564e-07,
"loss": 0.0017,
"step": 5025
},
{
"epoch": 17.81,
"learning_rate": 7.668205128205128e-07,
"loss": 0.003,
"step": 5050
},
{
"epoch": 17.9,
"learning_rate": 7.655384615384615e-07,
"loss": 0.0016,
"step": 5075
},
{
"epoch": 17.99,
"learning_rate": 7.642564102564102e-07,
"loss": 0.0018,
"step": 5100
},
{
"epoch": 18.08,
"learning_rate": 7.629743589743589e-07,
"loss": 0.0016,
"step": 5125
},
{
"epoch": 18.17,
"learning_rate": 7.616923076923077e-07,
"loss": 0.0011,
"step": 5150
},
{
"epoch": 18.25,
"learning_rate": 7.604102564102564e-07,
"loss": 0.0015,
"step": 5175
},
{
"epoch": 18.34,
"learning_rate": 7.591282051282051e-07,
"loss": 0.0011,
"step": 5200
},
{
"epoch": 18.43,
"learning_rate": 7.578461538461538e-07,
"loss": 0.002,
"step": 5225
},
{
"epoch": 18.52,
"learning_rate": 7.565641025641026e-07,
"loss": 0.0015,
"step": 5250
},
{
"epoch": 18.61,
"learning_rate": 7.552820512820513e-07,
"loss": 0.0019,
"step": 5275
},
{
"epoch": 18.69,
"learning_rate": 7.54e-07,
"loss": 0.0018,
"step": 5300
},
{
"epoch": 18.78,
"learning_rate": 7.527179487179487e-07,
"loss": 0.0013,
"step": 5325
},
{
"epoch": 18.87,
"learning_rate": 7.514358974358975e-07,
"loss": 0.0013,
"step": 5350
},
{
"epoch": 18.96,
"learning_rate": 7.501538461538462e-07,
"loss": 0.0017,
"step": 5375
},
{
"epoch": 19.05,
"learning_rate": 7.488717948717949e-07,
"loss": 0.0015,
"step": 5400
},
{
"epoch": 19.14,
"learning_rate": 7.475897435897436e-07,
"loss": 0.0012,
"step": 5425
},
{
"epoch": 19.22,
"learning_rate": 7.463076923076924e-07,
"loss": 0.0012,
"step": 5450
},
{
"epoch": 19.31,
"learning_rate": 7.450256410256411e-07,
"loss": 0.0011,
"step": 5475
},
{
"epoch": 19.4,
"learning_rate": 7.437435897435898e-07,
"loss": 0.0019,
"step": 5500
},
{
"epoch": 19.49,
"learning_rate": 7.424615384615385e-07,
"loss": 0.0018,
"step": 5525
},
{
"epoch": 19.58,
"learning_rate": 7.411794871794871e-07,
"loss": 0.0021,
"step": 5550
},
{
"epoch": 19.66,
"learning_rate": 7.398974358974359e-07,
"loss": 0.0021,
"step": 5575
},
{
"epoch": 19.75,
"learning_rate": 7.386153846153845e-07,
"loss": 0.0021,
"step": 5600
},
{
"epoch": 19.84,
"learning_rate": 7.373333333333332e-07,
"loss": 0.0014,
"step": 5625
},
{
"epoch": 19.93,
"learning_rate": 7.360512820512819e-07,
"loss": 0.0016,
"step": 5650
},
{
"epoch": 20.02,
"learning_rate": 7.347692307692307e-07,
"loss": 0.0011,
"step": 5675
},
{
"epoch": 20.11,
"learning_rate": 7.334871794871794e-07,
"loss": 0.0013,
"step": 5700
},
{
"epoch": 20.19,
"learning_rate": 7.322051282051281e-07,
"loss": 0.001,
"step": 5725
},
{
"epoch": 20.28,
"learning_rate": 7.309230769230768e-07,
"loss": 0.0009,
"step": 5750
},
{
"epoch": 20.37,
"learning_rate": 7.296410256410256e-07,
"loss": 0.0011,
"step": 5775
},
{
"epoch": 20.46,
"learning_rate": 7.283589743589743e-07,
"loss": 0.0023,
"step": 5800
},
{
"epoch": 20.55,
"learning_rate": 7.27076923076923e-07,
"loss": 0.002,
"step": 5825
},
{
"epoch": 20.63,
"learning_rate": 7.257948717948717e-07,
"loss": 0.0022,
"step": 5850
},
{
"epoch": 20.72,
"learning_rate": 7.245128205128205e-07,
"loss": 0.0012,
"step": 5875
},
{
"epoch": 20.81,
"learning_rate": 7.232307692307692e-07,
"loss": 0.001,
"step": 5900
},
{
"epoch": 20.9,
"learning_rate": 7.219487179487179e-07,
"loss": 0.0017,
"step": 5925
},
{
"epoch": 20.99,
"learning_rate": 7.206666666666666e-07,
"loss": 0.0018,
"step": 5950
},
{
"epoch": 21.08,
"learning_rate": 7.193846153846154e-07,
"loss": 0.0015,
"step": 5975
},
{
"epoch": 21.16,
"learning_rate": 7.181025641025641e-07,
"loss": 0.0027,
"step": 6000
},
{
"epoch": 21.16,
"eval_loss": 0.3372918367385864,
"eval_runtime": 2239.9939,
"eval_samples_per_second": 4.05,
"eval_steps_per_second": 0.253,
"eval_wer": 4.647915092169684,
"step": 6000
},
{
"epoch": 21.25,
"learning_rate": 7.168205128205128e-07,
"loss": 0.0013,
"step": 6025
},
{
"epoch": 21.34,
"learning_rate": 7.155384615384615e-07,
"loss": 0.001,
"step": 6050
},
{
"epoch": 21.43,
"learning_rate": 7.142564102564103e-07,
"loss": 0.0013,
"step": 6075
},
{
"epoch": 21.52,
"learning_rate": 7.12974358974359e-07,
"loss": 0.0013,
"step": 6100
},
{
"epoch": 21.6,
"learning_rate": 7.116923076923077e-07,
"loss": 0.0014,
"step": 6125
},
{
"epoch": 21.69,
"learning_rate": 7.104102564102564e-07,
"loss": 0.001,
"step": 6150
},
{
"epoch": 21.78,
"learning_rate": 7.091282051282052e-07,
"loss": 0.001,
"step": 6175
},
{
"epoch": 21.87,
"learning_rate": 7.078461538461538e-07,
"loss": 0.001,
"step": 6200
},
{
"epoch": 21.96,
"learning_rate": 7.065641025641025e-07,
"loss": 0.0015,
"step": 6225
},
{
"epoch": 22.05,
"learning_rate": 7.052820512820512e-07,
"loss": 0.0006,
"step": 6250
},
{
"epoch": 22.13,
"learning_rate": 7.04e-07,
"loss": 0.0011,
"step": 6275
},
{
"epoch": 22.22,
"learning_rate": 7.027179487179487e-07,
"loss": 0.0014,
"step": 6300
},
{
"epoch": 22.31,
"learning_rate": 7.014358974358974e-07,
"loss": 0.0009,
"step": 6325
},
{
"epoch": 22.4,
"learning_rate": 7.001538461538461e-07,
"loss": 0.0014,
"step": 6350
},
{
"epoch": 22.49,
"learning_rate": 6.988717948717949e-07,
"loss": 0.0009,
"step": 6375
},
{
"epoch": 22.57,
"learning_rate": 6.975897435897436e-07,
"loss": 0.0011,
"step": 6400
},
{
"epoch": 22.66,
"learning_rate": 6.963076923076923e-07,
"loss": 0.0009,
"step": 6425
},
{
"epoch": 22.75,
"learning_rate": 6.95025641025641e-07,
"loss": 0.0011,
"step": 6450
},
{
"epoch": 22.84,
"learning_rate": 6.937435897435897e-07,
"loss": 0.0009,
"step": 6475
},
{
"epoch": 22.93,
"learning_rate": 6.924615384615385e-07,
"loss": 0.0007,
"step": 6500
},
{
"epoch": 23.02,
"learning_rate": 6.911794871794872e-07,
"loss": 0.0013,
"step": 6525
},
{
"epoch": 23.1,
"learning_rate": 6.898974358974359e-07,
"loss": 0.0012,
"step": 6550
},
{
"epoch": 23.19,
"learning_rate": 6.886153846153846e-07,
"loss": 0.0009,
"step": 6575
},
{
"epoch": 23.28,
"learning_rate": 6.873333333333334e-07,
"loss": 0.0006,
"step": 6600
},
{
"epoch": 23.37,
"learning_rate": 6.86051282051282e-07,
"loss": 0.0015,
"step": 6625
},
{
"epoch": 23.46,
"learning_rate": 6.847692307692307e-07,
"loss": 0.0015,
"step": 6650
},
{
"epoch": 23.54,
"learning_rate": 6.834871794871794e-07,
"loss": 0.0011,
"step": 6675
},
{
"epoch": 23.63,
"learning_rate": 6.822051282051282e-07,
"loss": 0.0008,
"step": 6700
},
{
"epoch": 23.72,
"learning_rate": 6.809230769230769e-07,
"loss": 0.0009,
"step": 6725
},
{
"epoch": 23.81,
"learning_rate": 6.796410256410256e-07,
"loss": 0.001,
"step": 6750
},
{
"epoch": 23.9,
"learning_rate": 6.783589743589743e-07,
"loss": 0.0015,
"step": 6775
},
{
"epoch": 23.99,
"learning_rate": 6.770769230769231e-07,
"loss": 0.0009,
"step": 6800
},
{
"epoch": 24.07,
"learning_rate": 6.757948717948717e-07,
"loss": 0.0009,
"step": 6825
},
{
"epoch": 24.16,
"learning_rate": 6.745128205128204e-07,
"loss": 0.0007,
"step": 6850
},
{
"epoch": 24.25,
"learning_rate": 6.732307692307691e-07,
"loss": 0.0007,
"step": 6875
},
{
"epoch": 24.34,
"learning_rate": 6.72051282051282e-07,
"loss": 0.0119,
"step": 6900
},
{
"epoch": 24.43,
"learning_rate": 6.707692307692307e-07,
"loss": 0.001,
"step": 6925
},
{
"epoch": 24.51,
"learning_rate": 6.694871794871794e-07,
"loss": 0.0006,
"step": 6950
},
{
"epoch": 24.6,
"learning_rate": 6.682051282051282e-07,
"loss": 0.0006,
"step": 6975
},
{
"epoch": 24.69,
"learning_rate": 6.669230769230769e-07,
"loss": 0.0009,
"step": 7000
},
{
"epoch": 24.69,
"eval_loss": 0.3642103970050812,
"eval_runtime": 2265.8844,
"eval_samples_per_second": 4.004,
"eval_steps_per_second": 0.25,
"eval_wer": 4.708704366970065,
"step": 7000
},
{
"epoch": 24.78,
"learning_rate": 6.656410256410256e-07,
"loss": 0.001,
"step": 7025
},
{
"epoch": 24.87,
"learning_rate": 6.643589743589743e-07,
"loss": 0.0009,
"step": 7050
},
{
"epoch": 24.96,
"learning_rate": 6.630769230769231e-07,
"loss": 0.0026,
"step": 7075
},
{
"epoch": 25.04,
"learning_rate": 6.617948717948718e-07,
"loss": 0.0008,
"step": 7100
},
{
"epoch": 25.13,
"learning_rate": 6.605128205128205e-07,
"loss": 0.0007,
"step": 7125
},
{
"epoch": 25.22,
"learning_rate": 6.592307692307691e-07,
"loss": 0.0006,
"step": 7150
},
{
"epoch": 25.31,
"learning_rate": 6.579487179487179e-07,
"loss": 0.0008,
"step": 7175
},
{
"epoch": 25.4,
"learning_rate": 6.566666666666666e-07,
"loss": 0.0008,
"step": 7200
},
{
"epoch": 25.49,
"learning_rate": 6.553846153846153e-07,
"loss": 0.0008,
"step": 7225
},
{
"epoch": 25.57,
"learning_rate": 6.54102564102564e-07,
"loss": 0.0026,
"step": 7250
},
{
"epoch": 25.66,
"learning_rate": 6.528205128205128e-07,
"loss": 0.0009,
"step": 7275
},
{
"epoch": 25.75,
"learning_rate": 6.515384615384615e-07,
"loss": 0.0015,
"step": 7300
},
{
"epoch": 25.84,
"learning_rate": 6.502564102564102e-07,
"loss": 0.0006,
"step": 7325
},
{
"epoch": 25.93,
"learning_rate": 6.489743589743589e-07,
"loss": 0.0012,
"step": 7350
},
{
"epoch": 26.01,
"learning_rate": 6.476923076923077e-07,
"loss": 0.0009,
"step": 7375
},
{
"epoch": 26.1,
"learning_rate": 6.464102564102564e-07,
"loss": 0.0013,
"step": 7400
},
{
"epoch": 26.19,
"learning_rate": 6.451282051282051e-07,
"loss": 0.0011,
"step": 7425
},
{
"epoch": 26.28,
"learning_rate": 6.438461538461538e-07,
"loss": 0.0017,
"step": 7450
},
{
"epoch": 26.37,
"learning_rate": 6.425641025641026e-07,
"loss": 0.0014,
"step": 7475
},
{
"epoch": 26.46,
"learning_rate": 6.412820512820513e-07,
"loss": 0.0013,
"step": 7500
},
{
"epoch": 26.54,
"learning_rate": 6.4e-07,
"loss": 0.0013,
"step": 7525
},
{
"epoch": 26.63,
"learning_rate": 6.387179487179487e-07,
"loss": 0.0017,
"step": 7550
},
{
"epoch": 26.72,
"learning_rate": 6.374358974358975e-07,
"loss": 0.001,
"step": 7575
},
{
"epoch": 26.81,
"learning_rate": 6.361538461538462e-07,
"loss": 0.0012,
"step": 7600
},
{
"epoch": 26.9,
"learning_rate": 6.348717948717949e-07,
"loss": 0.0013,
"step": 7625
},
{
"epoch": 26.98,
"learning_rate": 6.335897435897436e-07,
"loss": 0.0008,
"step": 7650
},
{
"epoch": 27.07,
"learning_rate": 6.323076923076924e-07,
"loss": 0.0021,
"step": 7675
},
{
"epoch": 27.16,
"learning_rate": 6.310256410256411e-07,
"loss": 0.0007,
"step": 7700
},
{
"epoch": 27.25,
"learning_rate": 6.297435897435898e-07,
"loss": 0.001,
"step": 7725
},
{
"epoch": 27.34,
"learning_rate": 6.284615384615385e-07,
"loss": 0.0007,
"step": 7750
},
{
"epoch": 27.43,
"learning_rate": 6.27179487179487e-07,
"loss": 0.0012,
"step": 7775
},
{
"epoch": 27.51,
"learning_rate": 6.258974358974358e-07,
"loss": 0.0006,
"step": 7800
},
{
"epoch": 27.6,
"learning_rate": 6.246153846153845e-07,
"loss": 0.0006,
"step": 7825
},
{
"epoch": 27.69,
"learning_rate": 6.233333333333332e-07,
"loss": 0.0006,
"step": 7850
},
{
"epoch": 27.78,
"learning_rate": 6.22051282051282e-07,
"loss": 0.0005,
"step": 7875
},
{
"epoch": 27.87,
"learning_rate": 6.207692307692307e-07,
"loss": 0.0006,
"step": 7900
},
{
"epoch": 27.95,
"learning_rate": 6.194871794871794e-07,
"loss": 0.0008,
"step": 7925
},
{
"epoch": 28.04,
"learning_rate": 6.182051282051281e-07,
"loss": 0.0007,
"step": 7950
},
{
"epoch": 28.13,
"learning_rate": 6.169230769230768e-07,
"loss": 0.0004,
"step": 7975
},
{
"epoch": 28.22,
"learning_rate": 6.156410256410256e-07,
"loss": 0.0007,
"step": 8000
},
{
"epoch": 28.22,
"eval_loss": 0.3550758361816406,
"eval_runtime": 2263.6452,
"eval_samples_per_second": 4.008,
"eval_steps_per_second": 0.25,
"eval_wer": 4.661058719153551,
"step": 8000
},
{
"epoch": 28.31,
"learning_rate": 6.143589743589743e-07,
"loss": 0.0007,
"step": 8025
},
{
"epoch": 28.4,
"learning_rate": 6.13076923076923e-07,
"loss": 0.0013,
"step": 8050
},
{
"epoch": 28.48,
"learning_rate": 6.117948717948717e-07,
"loss": 0.0006,
"step": 8075
},
{
"epoch": 28.57,
"learning_rate": 6.105128205128205e-07,
"loss": 0.0008,
"step": 8100
},
{
"epoch": 28.66,
"learning_rate": 6.092307692307692e-07,
"loss": 0.0007,
"step": 8125
},
{
"epoch": 28.75,
"learning_rate": 6.079487179487179e-07,
"loss": 0.0004,
"step": 8150
},
{
"epoch": 28.84,
"learning_rate": 6.066666666666666e-07,
"loss": 0.0004,
"step": 8175
},
{
"epoch": 28.92,
"learning_rate": 6.053846153846154e-07,
"loss": 0.0011,
"step": 8200
},
{
"epoch": 29.01,
"learning_rate": 6.041025641025641e-07,
"loss": 0.0006,
"step": 8225
},
{
"epoch": 29.1,
"learning_rate": 6.028205128205128e-07,
"loss": 0.0005,
"step": 8250
},
{
"epoch": 29.19,
"learning_rate": 6.015384615384615e-07,
"loss": 0.0004,
"step": 8275
},
{
"epoch": 29.28,
"learning_rate": 6.002564102564103e-07,
"loss": 0.0009,
"step": 8300
},
{
"epoch": 29.37,
"learning_rate": 5.98974358974359e-07,
"loss": 0.0004,
"step": 8325
},
{
"epoch": 29.45,
"learning_rate": 5.976923076923077e-07,
"loss": 0.0004,
"step": 8350
},
{
"epoch": 29.54,
"learning_rate": 5.964102564102563e-07,
"loss": 0.0004,
"step": 8375
},
{
"epoch": 29.63,
"learning_rate": 5.951282051282051e-07,
"loss": 0.0005,
"step": 8400
},
{
"epoch": 29.72,
"learning_rate": 5.938461538461538e-07,
"loss": 0.0008,
"step": 8425
},
{
"epoch": 29.81,
"learning_rate": 5.925641025641025e-07,
"loss": 0.0007,
"step": 8450
},
{
"epoch": 29.89,
"learning_rate": 5.912820512820512e-07,
"loss": 0.0006,
"step": 8475
},
{
"epoch": 29.98,
"learning_rate": 5.9e-07,
"loss": 0.0026,
"step": 8500
},
{
"epoch": 30.07,
"learning_rate": 5.887179487179487e-07,
"loss": 0.0006,
"step": 8525
},
{
"epoch": 30.16,
"learning_rate": 5.874358974358974e-07,
"loss": 0.0012,
"step": 8550
},
{
"epoch": 30.25,
"learning_rate": 5.861538461538461e-07,
"loss": 0.001,
"step": 8575
},
{
"epoch": 30.34,
"learning_rate": 5.848717948717949e-07,
"loss": 0.0009,
"step": 8600
},
{
"epoch": 30.42,
"learning_rate": 5.835897435897436e-07,
"loss": 0.0006,
"step": 8625
},
{
"epoch": 30.51,
"learning_rate": 5.823076923076923e-07,
"loss": 0.0007,
"step": 8650
},
{
"epoch": 30.6,
"learning_rate": 5.81025641025641e-07,
"loss": 0.0007,
"step": 8675
},
{
"epoch": 30.69,
"learning_rate": 5.797435897435898e-07,
"loss": 0.0024,
"step": 8700
},
{
"epoch": 30.78,
"learning_rate": 5.784615384615385e-07,
"loss": 0.0017,
"step": 8725
},
{
"epoch": 30.86,
"learning_rate": 5.771794871794872e-07,
"loss": 0.0012,
"step": 8750
},
{
"epoch": 30.95,
"learning_rate": 5.758974358974359e-07,
"loss": 0.0012,
"step": 8775
},
{
"epoch": 31.04,
"learning_rate": 5.746153846153847e-07,
"loss": 0.0007,
"step": 8800
},
{
"epoch": 31.13,
"learning_rate": 5.733333333333334e-07,
"loss": 0.0008,
"step": 8825
},
{
"epoch": 31.22,
"learning_rate": 5.720512820512821e-07,
"loss": 0.0013,
"step": 8850
},
{
"epoch": 31.31,
"learning_rate": 5.707692307692307e-07,
"loss": 0.0009,
"step": 8875
},
{
"epoch": 31.39,
"learning_rate": 5.694871794871795e-07,
"loss": 0.0009,
"step": 8900
},
{
"epoch": 31.48,
"learning_rate": 5.682051282051282e-07,
"loss": 0.001,
"step": 8925
},
{
"epoch": 31.57,
"learning_rate": 5.669230769230769e-07,
"loss": 0.0005,
"step": 8950
},
{
"epoch": 31.66,
"learning_rate": 5.656410256410256e-07,
"loss": 0.0004,
"step": 8975
},
{
"epoch": 31.75,
"learning_rate": 5.643589743589742e-07,
"loss": 0.0006,
"step": 9000
},
{
"epoch": 31.75,
"eval_loss": 0.37413859367370605,
"eval_runtime": 2262.284,
"eval_samples_per_second": 4.01,
"eval_steps_per_second": 0.251,
"eval_wer": 4.748135247921664,
"step": 9000
},
{
"epoch": 31.83,
"learning_rate": 5.63076923076923e-07,
"loss": 0.0007,
"step": 9025
},
{
"epoch": 31.92,
"learning_rate": 5.617948717948717e-07,
"loss": 0.0013,
"step": 9050
},
{
"epoch": 32.01,
"learning_rate": 5.605128205128204e-07,
"loss": 0.0006,
"step": 9075
},
{
"epoch": 32.1,
"learning_rate": 5.592307692307691e-07,
"loss": 0.0006,
"step": 9100
},
{
"epoch": 32.19,
"learning_rate": 5.579487179487179e-07,
"loss": 0.0004,
"step": 9125
},
{
"epoch": 32.28,
"learning_rate": 5.566666666666666e-07,
"loss": 0.0007,
"step": 9150
},
{
"epoch": 32.36,
"learning_rate": 5.553846153846153e-07,
"loss": 0.0005,
"step": 9175
},
{
"epoch": 32.45,
"learning_rate": 5.54102564102564e-07,
"loss": 0.0003,
"step": 9200
},
{
"epoch": 32.54,
"learning_rate": 5.528205128205128e-07,
"loss": 0.0005,
"step": 9225
},
{
"epoch": 32.63,
"learning_rate": 5.515384615384615e-07,
"loss": 0.0008,
"step": 9250
},
{
"epoch": 32.72,
"learning_rate": 5.502564102564102e-07,
"loss": 0.0007,
"step": 9275
},
{
"epoch": 32.8,
"learning_rate": 5.489743589743589e-07,
"loss": 0.0008,
"step": 9300
},
{
"epoch": 32.89,
"learning_rate": 5.476923076923077e-07,
"loss": 0.0017,
"step": 9325
},
{
"epoch": 32.98,
"learning_rate": 5.464102564102564e-07,
"loss": 0.0016,
"step": 9350
},
{
"epoch": 33.07,
"learning_rate": 5.451282051282051e-07,
"loss": 0.0005,
"step": 9375
},
{
"epoch": 33.16,
"learning_rate": 5.438461538461538e-07,
"loss": 0.0005,
"step": 9400
},
{
"epoch": 33.25,
"learning_rate": 5.425641025641026e-07,
"loss": 0.0004,
"step": 9425
},
{
"epoch": 33.33,
"learning_rate": 5.412820512820513e-07,
"loss": 0.0005,
"step": 9450
},
{
"epoch": 33.42,
"learning_rate": 5.4e-07,
"loss": 0.0004,
"step": 9475
},
{
"epoch": 33.51,
"learning_rate": 5.387179487179487e-07,
"loss": 0.0003,
"step": 9500
},
{
"epoch": 33.6,
"learning_rate": 5.374358974358975e-07,
"loss": 0.0014,
"step": 9525
},
{
"epoch": 33.69,
"learning_rate": 5.361538461538462e-07,
"loss": 0.0006,
"step": 9550
},
{
"epoch": 33.77,
"learning_rate": 5.348717948717949e-07,
"loss": 0.0009,
"step": 9575
},
{
"epoch": 33.86,
"learning_rate": 5.335897435897435e-07,
"loss": 0.001,
"step": 9600
},
{
"epoch": 33.95,
"learning_rate": 5.323076923076923e-07,
"loss": 0.0005,
"step": 9625
},
{
"epoch": 34.04,
"learning_rate": 5.31025641025641e-07,
"loss": 0.0004,
"step": 9650
},
{
"epoch": 34.13,
"learning_rate": 5.297435897435897e-07,
"loss": 0.0004,
"step": 9675
},
{
"epoch": 34.22,
"learning_rate": 5.284615384615384e-07,
"loss": 0.0005,
"step": 9700
},
{
"epoch": 34.3,
"learning_rate": 5.271794871794872e-07,
"loss": 0.0004,
"step": 9725
},
{
"epoch": 34.39,
"learning_rate": 5.258974358974359e-07,
"loss": 0.0006,
"step": 9750
},
{
"epoch": 34.48,
"learning_rate": 5.246153846153846e-07,
"loss": 0.0007,
"step": 9775
},
{
"epoch": 34.57,
"learning_rate": 5.233333333333333e-07,
"loss": 0.0006,
"step": 9800
},
{
"epoch": 34.66,
"learning_rate": 5.220512820512821e-07,
"loss": 0.0005,
"step": 9825
},
{
"epoch": 34.74,
"learning_rate": 5.207692307692308e-07,
"loss": 0.0007,
"step": 9850
},
{
"epoch": 34.83,
"learning_rate": 5.194871794871795e-07,
"loss": 0.0005,
"step": 9875
},
{
"epoch": 34.92,
"learning_rate": 5.182051282051282e-07,
"loss": 0.0003,
"step": 9900
},
{
"epoch": 35.01,
"learning_rate": 5.16923076923077e-07,
"loss": 0.0006,
"step": 9925
},
{
"epoch": 35.1,
"learning_rate": 5.156410256410256e-07,
"loss": 0.0004,
"step": 9950
},
{
"epoch": 35.19,
"learning_rate": 5.143589743589743e-07,
"loss": 0.0005,
"step": 9975
},
{
"epoch": 35.27,
"learning_rate": 5.13076923076923e-07,
"loss": 0.0004,
"step": 10000
},
{
"epoch": 35.27,
"eval_loss": 0.37553662061691284,
"eval_runtime": 2266.5672,
"eval_samples_per_second": 4.003,
"eval_steps_per_second": 0.25,
"eval_wer": 4.679131206256367,
"step": 10000
},
{
"epoch": 35.36,
"learning_rate": 5.117948717948718e-07,
"loss": 0.0005,
"step": 10025
},
{
"epoch": 35.45,
"learning_rate": 5.105128205128205e-07,
"loss": 0.0007,
"step": 10050
},
{
"epoch": 35.54,
"learning_rate": 5.092307692307692e-07,
"loss": 0.0005,
"step": 10075
},
{
"epoch": 35.63,
"learning_rate": 5.079487179487179e-07,
"loss": 0.0004,
"step": 10100
},
{
"epoch": 35.71,
"learning_rate": 5.066666666666667e-07,
"loss": 0.0005,
"step": 10125
},
{
"epoch": 35.8,
"learning_rate": 5.053846153846154e-07,
"loss": 0.0004,
"step": 10150
},
{
"epoch": 35.89,
"learning_rate": 5.041025641025641e-07,
"loss": 0.0003,
"step": 10175
},
{
"epoch": 35.98,
"learning_rate": 5.028205128205127e-07,
"loss": 0.0003,
"step": 10200
},
{
"epoch": 36.07,
"learning_rate": 5.015384615384614e-07,
"loss": 0.0005,
"step": 10225
},
{
"epoch": 36.16,
"learning_rate": 5.002564102564102e-07,
"loss": 0.0005,
"step": 10250
},
{
"epoch": 36.24,
"learning_rate": 4.98974358974359e-07,
"loss": 0.0008,
"step": 10275
},
{
"epoch": 36.33,
"learning_rate": 4.976923076923077e-07,
"loss": 0.0004,
"step": 10300
},
{
"epoch": 36.42,
"learning_rate": 4.964102564102564e-07,
"loss": 0.0003,
"step": 10325
},
{
"epoch": 36.51,
"learning_rate": 4.951282051282051e-07,
"loss": 0.0006,
"step": 10350
},
{
"epoch": 36.6,
"learning_rate": 4.938461538461538e-07,
"loss": 0.0006,
"step": 10375
},
{
"epoch": 36.68,
"learning_rate": 4.925641025641025e-07,
"loss": 0.0003,
"step": 10400
},
{
"epoch": 36.77,
"learning_rate": 4.912820512820512e-07,
"loss": 0.0003,
"step": 10425
},
{
"epoch": 36.86,
"learning_rate": 4.9e-07,
"loss": 0.0004,
"step": 10450
},
{
"epoch": 36.95,
"learning_rate": 4.887179487179487e-07,
"loss": 0.0004,
"step": 10475
},
{
"epoch": 37.04,
"learning_rate": 4.874358974358974e-07,
"loss": 0.0003,
"step": 10500
},
{
"epoch": 37.13,
"learning_rate": 4.861538461538461e-07,
"loss": 0.0003,
"step": 10525
},
{
"epoch": 37.21,
"learning_rate": 4.848717948717949e-07,
"loss": 0.0003,
"step": 10550
},
{
"epoch": 37.3,
"learning_rate": 4.835897435897436e-07,
"loss": 0.0003,
"step": 10575
},
{
"epoch": 37.39,
"learning_rate": 4.823076923076923e-07,
"loss": 0.0003,
"step": 10600
},
{
"epoch": 37.48,
"learning_rate": 4.81025641025641e-07,
"loss": 0.0002,
"step": 10625
},
{
"epoch": 37.57,
"learning_rate": 4.797435897435898e-07,
"loss": 0.0002,
"step": 10650
},
{
"epoch": 37.65,
"learning_rate": 4.784615384615384e-07,
"loss": 0.0006,
"step": 10675
},
{
"epoch": 37.74,
"learning_rate": 4.771794871794871e-07,
"loss": 0.0004,
"step": 10700
},
{
"epoch": 37.83,
"learning_rate": 4.758974358974359e-07,
"loss": 0.0011,
"step": 10725
},
{
"epoch": 37.92,
"learning_rate": 4.7461538461538456e-07,
"loss": 0.0007,
"step": 10750
},
{
"epoch": 38.01,
"learning_rate": 4.733333333333333e-07,
"loss": 0.0008,
"step": 10775
},
{
"epoch": 38.1,
"learning_rate": 4.72051282051282e-07,
"loss": 0.0007,
"step": 10800
},
{
"epoch": 38.18,
"learning_rate": 4.7076923076923073e-07,
"loss": 0.0012,
"step": 10825
},
{
"epoch": 38.27,
"learning_rate": 4.6948717948717946e-07,
"loss": 0.0012,
"step": 10850
},
{
"epoch": 38.36,
"learning_rate": 4.682051282051282e-07,
"loss": 0.0008,
"step": 10875
},
{
"epoch": 38.45,
"learning_rate": 4.669230769230769e-07,
"loss": 0.0005,
"step": 10900
},
{
"epoch": 38.54,
"learning_rate": 4.6564102564102563e-07,
"loss": 0.0004,
"step": 10925
},
{
"epoch": 38.62,
"learning_rate": 4.6435897435897436e-07,
"loss": 0.0009,
"step": 10950
},
{
"epoch": 38.71,
"learning_rate": 4.6307692307692303e-07,
"loss": 0.0004,
"step": 10975
},
{
"epoch": 38.8,
"learning_rate": 4.6179487179487175e-07,
"loss": 0.0008,
"step": 11000
},
{
"epoch": 38.8,
"eval_loss": 0.3690282702445984,
"eval_runtime": 2259.4792,
"eval_samples_per_second": 4.015,
"eval_steps_per_second": 0.251,
"eval_wer": 4.638057371931784,
"step": 11000
},
{
"epoch": 38.89,
"learning_rate": 4.605128205128205e-07,
"loss": 0.0006,
"step": 11025
},
{
"epoch": 38.98,
"learning_rate": 4.592307692307692e-07,
"loss": 0.0007,
"step": 11050
},
{
"epoch": 39.07,
"learning_rate": 4.5794871794871793e-07,
"loss": 0.0003,
"step": 11075
},
{
"epoch": 39.15,
"learning_rate": 4.5666666666666665e-07,
"loss": 0.0006,
"step": 11100
},
{
"epoch": 39.24,
"learning_rate": 4.553846153846154e-07,
"loss": 0.0005,
"step": 11125
},
{
"epoch": 39.33,
"learning_rate": 4.541025641025641e-07,
"loss": 0.0003,
"step": 11150
},
{
"epoch": 39.42,
"learning_rate": 4.5282051282051283e-07,
"loss": 0.0004,
"step": 11175
},
{
"epoch": 39.51,
"learning_rate": 4.5153846153846155e-07,
"loss": 0.0006,
"step": 11200
},
{
"epoch": 39.59,
"learning_rate": 4.502564102564103e-07,
"loss": 0.0003,
"step": 11225
},
{
"epoch": 39.68,
"learning_rate": 4.48974358974359e-07,
"loss": 0.0004,
"step": 11250
},
{
"epoch": 39.77,
"learning_rate": 4.476923076923076e-07,
"loss": 0.0006,
"step": 11275
},
{
"epoch": 39.86,
"learning_rate": 4.4641025641025635e-07,
"loss": 0.0004,
"step": 11300
},
{
"epoch": 39.95,
"learning_rate": 4.4512820512820507e-07,
"loss": 0.0004,
"step": 11325
},
{
"epoch": 40.04,
"learning_rate": 4.438461538461538e-07,
"loss": 0.0011,
"step": 11350
},
{
"epoch": 40.12,
"learning_rate": 4.425641025641025e-07,
"loss": 0.0003,
"step": 11375
},
{
"epoch": 40.21,
"learning_rate": 4.4128205128205124e-07,
"loss": 0.0002,
"step": 11400
},
{
"epoch": 40.3,
"learning_rate": 4.3999999999999997e-07,
"loss": 0.0002,
"step": 11425
},
{
"epoch": 40.39,
"learning_rate": 4.387179487179487e-07,
"loss": 0.0002,
"step": 11450
},
{
"epoch": 40.48,
"learning_rate": 4.374358974358974e-07,
"loss": 0.0002,
"step": 11475
},
{
"epoch": 40.56,
"learning_rate": 4.3615384615384614e-07,
"loss": 0.0002,
"step": 11500
},
{
"epoch": 40.65,
"learning_rate": 4.3487179487179487e-07,
"loss": 0.0003,
"step": 11525
},
{
"epoch": 40.74,
"learning_rate": 4.335897435897436e-07,
"loss": 0.0002,
"step": 11550
},
{
"epoch": 40.83,
"learning_rate": 4.3230769230769227e-07,
"loss": 0.0002,
"step": 11575
},
{
"epoch": 40.92,
"learning_rate": 4.31025641025641e-07,
"loss": 0.0002,
"step": 11600
},
{
"epoch": 41.01,
"learning_rate": 4.297435897435897e-07,
"loss": 0.0009,
"step": 11625
},
{
"epoch": 41.09,
"learning_rate": 4.2846153846153844e-07,
"loss": 0.0008,
"step": 11650
},
{
"epoch": 41.18,
"learning_rate": 4.2717948717948716e-07,
"loss": 0.0002,
"step": 11675
},
{
"epoch": 41.27,
"learning_rate": 4.258974358974359e-07,
"loss": 0.0002,
"step": 11700
},
{
"epoch": 41.36,
"learning_rate": 4.246153846153846e-07,
"loss": 0.0002,
"step": 11725
},
{
"epoch": 41.45,
"learning_rate": 4.2333333333333334e-07,
"loss": 0.0002,
"step": 11750
},
{
"epoch": 41.53,
"learning_rate": 4.2205128205128206e-07,
"loss": 0.0002,
"step": 11775
},
{
"epoch": 41.62,
"learning_rate": 4.207692307692308e-07,
"loss": 0.0002,
"step": 11800
},
{
"epoch": 41.71,
"learning_rate": 4.194871794871795e-07,
"loss": 0.0004,
"step": 11825
},
{
"epoch": 41.8,
"learning_rate": 4.182051282051282e-07,
"loss": 0.0002,
"step": 11850
},
{
"epoch": 41.89,
"learning_rate": 4.1692307692307686e-07,
"loss": 0.0003,
"step": 11875
},
{
"epoch": 41.98,
"learning_rate": 4.156410256410256e-07,
"loss": 0.0002,
"step": 11900
},
{
"epoch": 42.06,
"learning_rate": 4.143589743589743e-07,
"loss": 0.0002,
"step": 11925
},
{
"epoch": 42.15,
"learning_rate": 4.1307692307692303e-07,
"loss": 0.0003,
"step": 11950
},
{
"epoch": 42.24,
"learning_rate": 4.1179487179487176e-07,
"loss": 0.0002,
"step": 11975
},
{
"epoch": 42.33,
"learning_rate": 4.105128205128205e-07,
"loss": 0.0002,
"step": 12000
},
{
"epoch": 42.33,
"eval_loss": 0.38876286149024963,
"eval_runtime": 2257.1594,
"eval_samples_per_second": 4.019,
"eval_steps_per_second": 0.251,
"eval_wer": 4.511549962212072,
"step": 12000
},
{
"epoch": 42.42,
"learning_rate": 4.092307692307692e-07,
"loss": 0.0002,
"step": 12025
},
{
"epoch": 42.5,
"learning_rate": 4.0794871794871793e-07,
"loss": 0.0002,
"step": 12050
},
{
"epoch": 42.59,
"learning_rate": 4.0666666666666666e-07,
"loss": 0.0004,
"step": 12075
},
{
"epoch": 42.68,
"learning_rate": 4.053846153846154e-07,
"loss": 0.0002,
"step": 12100
},
{
"epoch": 42.77,
"learning_rate": 4.041025641025641e-07,
"loss": 0.0002,
"step": 12125
},
{
"epoch": 42.86,
"learning_rate": 4.0282051282051283e-07,
"loss": 0.0002,
"step": 12150
},
{
"epoch": 42.95,
"learning_rate": 4.015384615384615e-07,
"loss": 0.0002,
"step": 12175
},
{
"epoch": 43.03,
"learning_rate": 4.002564102564102e-07,
"loss": 0.0002,
"step": 12200
},
{
"epoch": 43.12,
"learning_rate": 3.9897435897435895e-07,
"loss": 0.0002,
"step": 12225
},
{
"epoch": 43.21,
"learning_rate": 3.976923076923077e-07,
"loss": 0.0002,
"step": 12250
},
{
"epoch": 43.3,
"learning_rate": 3.964102564102564e-07,
"loss": 0.0002,
"step": 12275
},
{
"epoch": 43.39,
"learning_rate": 3.951282051282051e-07,
"loss": 0.0002,
"step": 12300
},
{
"epoch": 43.47,
"learning_rate": 3.9384615384615385e-07,
"loss": 0.0002,
"step": 12325
},
{
"epoch": 43.56,
"learning_rate": 3.925641025641026e-07,
"loss": 0.0002,
"step": 12350
},
{
"epoch": 43.65,
"learning_rate": 3.9128205128205125e-07,
"loss": 0.0002,
"step": 12375
},
{
"epoch": 43.74,
"learning_rate": 3.8999999999999997e-07,
"loss": 0.0002,
"step": 12400
},
{
"epoch": 43.83,
"learning_rate": 3.887179487179487e-07,
"loss": 0.0007,
"step": 12425
},
{
"epoch": 43.92,
"learning_rate": 3.874358974358974e-07,
"loss": 0.0002,
"step": 12450
},
{
"epoch": 44.0,
"learning_rate": 3.8615384615384615e-07,
"loss": 0.0002,
"step": 12475
},
{
"epoch": 44.09,
"learning_rate": 3.848717948717948e-07,
"loss": 0.0002,
"step": 12500
},
{
"epoch": 44.18,
"learning_rate": 3.8358974358974354e-07,
"loss": 0.0002,
"step": 12525
},
{
"epoch": 44.27,
"learning_rate": 3.8230769230769227e-07,
"loss": 0.0002,
"step": 12550
},
{
"epoch": 44.36,
"learning_rate": 3.81025641025641e-07,
"loss": 0.0002,
"step": 12575
},
{
"epoch": 44.44,
"learning_rate": 3.797435897435897e-07,
"loss": 0.0002,
"step": 12600
},
{
"epoch": 44.53,
"learning_rate": 3.7846153846153844e-07,
"loss": 0.0002,
"step": 12625
},
{
"epoch": 44.62,
"learning_rate": 3.7717948717948717e-07,
"loss": 0.0002,
"step": 12650
},
{
"epoch": 44.71,
"learning_rate": 3.758974358974359e-07,
"loss": 0.0002,
"step": 12675
},
{
"epoch": 44.8,
"learning_rate": 3.746153846153846e-07,
"loss": 0.0002,
"step": 12700
},
{
"epoch": 44.89,
"learning_rate": 3.7333333333333334e-07,
"loss": 0.0002,
"step": 12725
},
{
"epoch": 44.97,
"learning_rate": 3.7205128205128207e-07,
"loss": 0.0002,
"step": 12750
},
{
"epoch": 45.06,
"learning_rate": 3.707692307692308e-07,
"loss": 0.0002,
"step": 12775
},
{
"epoch": 45.15,
"learning_rate": 3.6948717948717946e-07,
"loss": 0.0002,
"step": 12800
},
{
"epoch": 45.24,
"learning_rate": 3.682051282051282e-07,
"loss": 0.0002,
"step": 12825
},
{
"epoch": 45.33,
"learning_rate": 3.669230769230769e-07,
"loss": 0.0002,
"step": 12850
},
{
"epoch": 45.41,
"learning_rate": 3.6564102564102564e-07,
"loss": 0.0002,
"step": 12875
},
{
"epoch": 45.5,
"learning_rate": 3.6435897435897436e-07,
"loss": 0.0002,
"step": 12900
},
{
"epoch": 45.59,
"learning_rate": 3.6307692307692303e-07,
"loss": 0.0002,
"step": 12925
},
{
"epoch": 45.68,
"learning_rate": 3.6179487179487176e-07,
"loss": 0.0002,
"step": 12950
},
{
"epoch": 45.77,
"learning_rate": 3.605128205128205e-07,
"loss": 0.0002,
"step": 12975
},
{
"epoch": 45.86,
"learning_rate": 3.592307692307692e-07,
"loss": 0.0002,
"step": 13000
},
{
"epoch": 45.86,
"eval_loss": 0.3982228934764862,
"eval_runtime": 2241.1668,
"eval_samples_per_second": 4.048,
"eval_steps_per_second": 0.253,
"eval_wer": 4.58548286399632,
"step": 13000
},
{
"epoch": 45.94,
"learning_rate": 3.5794871794871793e-07,
"loss": 0.0002,
"step": 13025
},
{
"epoch": 46.03,
"learning_rate": 3.5666666666666666e-07,
"loss": 0.0002,
"step": 13050
},
{
"epoch": 46.12,
"learning_rate": 3.553846153846154e-07,
"loss": 0.0002,
"step": 13075
},
{
"epoch": 46.21,
"learning_rate": 3.5410256410256405e-07,
"loss": 0.0002,
"step": 13100
},
{
"epoch": 46.3,
"learning_rate": 3.528205128205128e-07,
"loss": 0.0001,
"step": 13125
},
{
"epoch": 46.38,
"learning_rate": 3.515384615384615e-07,
"loss": 0.0001,
"step": 13150
},
{
"epoch": 46.47,
"learning_rate": 3.5025641025641023e-07,
"loss": 0.0001,
"step": 13175
},
{
"epoch": 46.56,
"learning_rate": 3.4897435897435895e-07,
"loss": 0.0002,
"step": 13200
},
{
"epoch": 46.65,
"learning_rate": 3.476923076923077e-07,
"loss": 0.0002,
"step": 13225
},
{
"epoch": 46.74,
"learning_rate": 3.464102564102564e-07,
"loss": 0.0002,
"step": 13250
},
{
"epoch": 46.83,
"learning_rate": 3.4512820512820513e-07,
"loss": 0.0002,
"step": 13275
},
{
"epoch": 46.91,
"learning_rate": 3.4384615384615385e-07,
"loss": 0.0002,
"step": 13300
},
{
"epoch": 47.0,
"learning_rate": 3.425641025641026e-07,
"loss": 0.0002,
"step": 13325
},
{
"epoch": 47.09,
"learning_rate": 3.412820512820513e-07,
"loss": 0.0001,
"step": 13350
},
{
"epoch": 47.18,
"learning_rate": 3.4000000000000003e-07,
"loss": 0.0001,
"step": 13375
},
{
"epoch": 47.27,
"learning_rate": 3.387179487179487e-07,
"loss": 0.0001,
"step": 13400
},
{
"epoch": 47.35,
"learning_rate": 3.374358974358974e-07,
"loss": 0.0001,
"step": 13425
},
{
"epoch": 47.44,
"learning_rate": 3.361538461538461e-07,
"loss": 0.0002,
"step": 13450
},
{
"epoch": 47.53,
"learning_rate": 3.348717948717948e-07,
"loss": 0.0001,
"step": 13475
},
{
"epoch": 47.62,
"learning_rate": 3.3358974358974355e-07,
"loss": 0.0001,
"step": 13500
},
{
"epoch": 47.71,
"learning_rate": 3.3230769230769227e-07,
"loss": 0.0001,
"step": 13525
},
{
"epoch": 47.8,
"learning_rate": 3.31025641025641e-07,
"loss": 0.0001,
"step": 13550
},
{
"epoch": 47.88,
"learning_rate": 3.297435897435897e-07,
"loss": 0.0001,
"step": 13575
},
{
"epoch": 47.97,
"learning_rate": 3.2846153846153844e-07,
"loss": 0.0001,
"step": 13600
},
{
"epoch": 48.06,
"learning_rate": 3.2717948717948717e-07,
"loss": 0.0002,
"step": 13625
},
{
"epoch": 48.15,
"learning_rate": 3.258974358974359e-07,
"loss": 0.0001,
"step": 13650
},
{
"epoch": 48.24,
"learning_rate": 3.246153846153846e-07,
"loss": 0.0001,
"step": 13675
},
{
"epoch": 48.32,
"learning_rate": 3.233333333333333e-07,
"loss": 0.0001,
"step": 13700
},
{
"epoch": 48.41,
"learning_rate": 3.22051282051282e-07,
"loss": 0.0001,
"step": 13725
},
{
"epoch": 48.5,
"learning_rate": 3.2076923076923074e-07,
"loss": 0.0001,
"step": 13750
},
{
"epoch": 48.59,
"learning_rate": 3.1948717948717946e-07,
"loss": 0.0001,
"step": 13775
},
{
"epoch": 48.68,
"learning_rate": 3.182051282051282e-07,
"loss": 0.0001,
"step": 13800
},
{
"epoch": 48.77,
"learning_rate": 3.169230769230769e-07,
"loss": 0.0001,
"step": 13825
},
{
"epoch": 48.85,
"learning_rate": 3.1564102564102564e-07,
"loss": 0.0001,
"step": 13850
},
{
"epoch": 48.94,
"learning_rate": 3.1435897435897436e-07,
"loss": 0.0001,
"step": 13875
},
{
"epoch": 49.03,
"learning_rate": 3.130769230769231e-07,
"loss": 0.0001,
"step": 13900
},
{
"epoch": 49.12,
"learning_rate": 3.117948717948718e-07,
"loss": 0.0001,
"step": 13925
},
{
"epoch": 49.21,
"learning_rate": 3.1051282051282054e-07,
"loss": 0.0001,
"step": 13950
},
{
"epoch": 49.29,
"learning_rate": 3.0923076923076926e-07,
"loss": 0.0001,
"step": 13975
},
{
"epoch": 49.38,
"learning_rate": 3.07948717948718e-07,
"loss": 0.0001,
"step": 14000
},
{
"epoch": 49.38,
"eval_loss": 0.4039895832538605,
"eval_runtime": 2249.3433,
"eval_samples_per_second": 4.033,
"eval_steps_per_second": 0.252,
"eval_wer": 4.6084842112180855,
"step": 14000
},
{
"epoch": 49.47,
"learning_rate": 3.066666666666666e-07,
"loss": 0.0001,
"step": 14025
},
{
"epoch": 49.56,
"learning_rate": 3.0538461538461533e-07,
"loss": 0.0001,
"step": 14050
},
{
"epoch": 49.65,
"learning_rate": 3.0410256410256406e-07,
"loss": 0.0001,
"step": 14075
},
{
"epoch": 49.74,
"learning_rate": 3.028205128205128e-07,
"loss": 0.0001,
"step": 14100
},
{
"epoch": 49.82,
"learning_rate": 3.015384615384615e-07,
"loss": 0.0001,
"step": 14125
},
{
"epoch": 49.91,
"learning_rate": 3.0025641025641023e-07,
"loss": 0.0001,
"step": 14150
},
{
"epoch": 50.0,
"learning_rate": 2.9897435897435896e-07,
"loss": 0.0001,
"step": 14175
},
{
"epoch": 50.09,
"learning_rate": 2.976923076923077e-07,
"loss": 0.0001,
"step": 14200
},
{
"epoch": 50.18,
"learning_rate": 2.964102564102564e-07,
"loss": 0.0001,
"step": 14225
},
{
"epoch": 50.26,
"learning_rate": 2.9512820512820513e-07,
"loss": 0.0001,
"step": 14250
},
{
"epoch": 50.35,
"learning_rate": 2.9384615384615385e-07,
"loss": 0.0001,
"step": 14275
},
{
"epoch": 50.44,
"learning_rate": 2.925641025641026e-07,
"loss": 0.0001,
"step": 14300
},
{
"epoch": 50.53,
"learning_rate": 2.9128205128205125e-07,
"loss": 0.0001,
"step": 14325
},
{
"epoch": 50.62,
"learning_rate": 2.9e-07,
"loss": 0.0001,
"step": 14350
},
{
"epoch": 50.71,
"learning_rate": 2.887179487179487e-07,
"loss": 0.0001,
"step": 14375
},
{
"epoch": 50.79,
"learning_rate": 2.874358974358974e-07,
"loss": 0.0001,
"step": 14400
},
{
"epoch": 50.88,
"learning_rate": 2.8615384615384615e-07,
"loss": 0.0001,
"step": 14425
},
{
"epoch": 50.97,
"learning_rate": 2.848717948717949e-07,
"loss": 0.0001,
"step": 14450
},
{
"epoch": 51.06,
"learning_rate": 2.835897435897436e-07,
"loss": 0.0001,
"step": 14475
},
{
"epoch": 51.15,
"learning_rate": 2.823076923076923e-07,
"loss": 0.0001,
"step": 14500
},
{
"epoch": 51.23,
"learning_rate": 2.8102564102564105e-07,
"loss": 0.0001,
"step": 14525
},
{
"epoch": 51.32,
"learning_rate": 2.797435897435897e-07,
"loss": 0.0001,
"step": 14550
},
{
"epoch": 51.41,
"learning_rate": 2.7846153846153845e-07,
"loss": 0.0001,
"step": 14575
},
{
"epoch": 51.5,
"learning_rate": 2.7717948717948717e-07,
"loss": 0.0001,
"step": 14600
},
{
"epoch": 51.59,
"learning_rate": 2.7589743589743584e-07,
"loss": 0.0001,
"step": 14625
},
{
"epoch": 51.68,
"learning_rate": 2.7461538461538457e-07,
"loss": 0.0001,
"step": 14650
},
{
"epoch": 51.76,
"learning_rate": 2.733333333333333e-07,
"loss": 0.0001,
"step": 14675
},
{
"epoch": 51.85,
"learning_rate": 2.72051282051282e-07,
"loss": 0.0001,
"step": 14700
},
{
"epoch": 51.94,
"learning_rate": 2.7076923076923074e-07,
"loss": 0.0001,
"step": 14725
},
{
"epoch": 52.03,
"learning_rate": 2.6948717948717947e-07,
"loss": 0.0001,
"step": 14750
},
{
"epoch": 52.12,
"learning_rate": 2.682051282051282e-07,
"loss": 0.0001,
"step": 14775
},
{
"epoch": 52.2,
"learning_rate": 2.669230769230769e-07,
"loss": 0.0001,
"step": 14800
},
{
"epoch": 52.29,
"learning_rate": 2.6564102564102564e-07,
"loss": 0.0001,
"step": 14825
},
{
"epoch": 52.38,
"learning_rate": 2.6435897435897437e-07,
"loss": 0.0001,
"step": 14850
},
{
"epoch": 52.47,
"learning_rate": 2.630769230769231e-07,
"loss": 0.0001,
"step": 14875
},
{
"epoch": 52.56,
"learning_rate": 2.617948717948718e-07,
"loss": 0.0001,
"step": 14900
},
{
"epoch": 52.65,
"learning_rate": 2.605128205128205e-07,
"loss": 0.0001,
"step": 14925
},
{
"epoch": 52.73,
"learning_rate": 2.592307692307692e-07,
"loss": 0.0001,
"step": 14950
},
{
"epoch": 52.82,
"learning_rate": 2.5794871794871794e-07,
"loss": 0.0001,
"step": 14975
},
{
"epoch": 52.91,
"learning_rate": 2.5666666666666666e-07,
"loss": 0.0001,
"step": 15000
},
{
"epoch": 52.91,
"eval_loss": 0.41004934906959534,
"eval_runtime": 2412.5571,
"eval_samples_per_second": 3.76,
"eval_steps_per_second": 0.235,
"eval_wer": 4.588768770742287,
"step": 15000
},
{
"epoch": 53.0,
"learning_rate": 2.553846153846154e-07,
"loss": 0.0001,
"step": 15025
},
{
"epoch": 53.09,
"learning_rate": 2.541025641025641e-07,
"loss": 0.0001,
"step": 15050
},
{
"epoch": 53.17,
"learning_rate": 2.5282051282051284e-07,
"loss": 0.0001,
"step": 15075
},
{
"epoch": 53.26,
"learning_rate": 2.515384615384615e-07,
"loss": 0.0001,
"step": 15100
},
{
"epoch": 53.35,
"learning_rate": 2.5025641025641023e-07,
"loss": 0.0001,
"step": 15125
},
{
"epoch": 53.44,
"learning_rate": 2.4897435897435896e-07,
"loss": 0.0001,
"step": 15150
},
{
"epoch": 53.53,
"learning_rate": 2.476923076923077e-07,
"loss": 0.0001,
"step": 15175
},
{
"epoch": 53.62,
"learning_rate": 2.464102564102564e-07,
"loss": 0.0001,
"step": 15200
},
{
"epoch": 53.7,
"learning_rate": 2.4512820512820513e-07,
"loss": 0.0001,
"step": 15225
},
{
"epoch": 53.79,
"learning_rate": 2.4384615384615386e-07,
"loss": 0.0001,
"step": 15250
},
{
"epoch": 53.88,
"learning_rate": 2.425641025641026e-07,
"loss": 0.0001,
"step": 15275
},
{
"epoch": 53.97,
"learning_rate": 2.4128205128205125e-07,
"loss": 0.0001,
"step": 15300
},
{
"epoch": 54.06,
"learning_rate": 2.4e-07,
"loss": 0.0001,
"step": 15325
},
{
"epoch": 54.14,
"learning_rate": 2.387179487179487e-07,
"loss": 0.0001,
"step": 15350
},
{
"epoch": 54.23,
"learning_rate": 2.3743589743589743e-07,
"loss": 0.0001,
"step": 15375
},
{
"epoch": 54.32,
"learning_rate": 2.3615384615384615e-07,
"loss": 0.0001,
"step": 15400
},
{
"epoch": 54.41,
"learning_rate": 2.3487179487179488e-07,
"loss": 0.0001,
"step": 15425
},
{
"epoch": 54.5,
"learning_rate": 2.3358974358974358e-07,
"loss": 0.0001,
"step": 15450
},
{
"epoch": 54.59,
"learning_rate": 2.323076923076923e-07,
"loss": 0.0001,
"step": 15475
},
{
"epoch": 54.67,
"learning_rate": 2.31025641025641e-07,
"loss": 0.0001,
"step": 15500
},
{
"epoch": 54.76,
"learning_rate": 2.2974358974358972e-07,
"loss": 0.0001,
"step": 15525
},
{
"epoch": 54.85,
"learning_rate": 2.2846153846153845e-07,
"loss": 0.0001,
"step": 15550
},
{
"epoch": 54.94,
"learning_rate": 2.2717948717948717e-07,
"loss": 0.0001,
"step": 15575
},
{
"epoch": 55.03,
"learning_rate": 2.258974358974359e-07,
"loss": 0.0001,
"step": 15600
},
{
"epoch": 55.11,
"learning_rate": 2.246153846153846e-07,
"loss": 0.0001,
"step": 15625
},
{
"epoch": 55.2,
"learning_rate": 2.2333333333333332e-07,
"loss": 0.0001,
"step": 15650
},
{
"epoch": 55.29,
"learning_rate": 2.2205128205128205e-07,
"loss": 0.0001,
"step": 15675
},
{
"epoch": 55.38,
"learning_rate": 2.2076923076923077e-07,
"loss": 0.0001,
"step": 15700
},
{
"epoch": 55.47,
"learning_rate": 2.194871794871795e-07,
"loss": 0.0001,
"step": 15725
},
{
"epoch": 55.56,
"learning_rate": 2.1820512820512822e-07,
"loss": 0.0001,
"step": 15750
},
{
"epoch": 55.64,
"learning_rate": 2.169230769230769e-07,
"loss": 0.0001,
"step": 15775
},
{
"epoch": 55.73,
"learning_rate": 2.1564102564102562e-07,
"loss": 0.0001,
"step": 15800
},
{
"epoch": 55.82,
"learning_rate": 2.1435897435897434e-07,
"loss": 0.0001,
"step": 15825
},
{
"epoch": 55.91,
"learning_rate": 2.1307692307692307e-07,
"loss": 0.0001,
"step": 15850
},
{
"epoch": 56.0,
"learning_rate": 2.117948717948718e-07,
"loss": 0.0001,
"step": 15875
},
{
"epoch": 56.08,
"learning_rate": 2.1051282051282052e-07,
"loss": 0.0001,
"step": 15900
},
{
"epoch": 56.17,
"learning_rate": 2.0923076923076921e-07,
"loss": 0.0001,
"step": 15925
},
{
"epoch": 56.26,
"learning_rate": 2.0794871794871794e-07,
"loss": 0.0001,
"step": 15950
},
{
"epoch": 56.35,
"learning_rate": 2.0666666666666666e-07,
"loss": 0.0001,
"step": 15975
},
{
"epoch": 56.44,
"learning_rate": 2.053846153846154e-07,
"loss": 0.0001,
"step": 16000
},
{
"epoch": 56.44,
"eval_loss": 0.4165339171886444,
"eval_runtime": 2341.3426,
"eval_samples_per_second": 3.875,
"eval_steps_per_second": 0.242,
"eval_wer": 4.587125817369303,
"step": 16000
},
{
"epoch": 56.53,
"learning_rate": 2.0410256410256411e-07,
"loss": 0.0001,
"step": 16025
},
{
"epoch": 56.61,
"learning_rate": 2.028205128205128e-07,
"loss": 0.0001,
"step": 16050
},
{
"epoch": 56.7,
"learning_rate": 2.015384615384615e-07,
"loss": 0.0001,
"step": 16075
},
{
"epoch": 56.79,
"learning_rate": 2.0025641025641024e-07,
"loss": 0.0001,
"step": 16100
},
{
"epoch": 56.88,
"learning_rate": 1.9897435897435896e-07,
"loss": 0.0001,
"step": 16125
},
{
"epoch": 56.97,
"learning_rate": 1.9769230769230768e-07,
"loss": 0.0001,
"step": 16150
},
{
"epoch": 57.05,
"learning_rate": 1.964102564102564e-07,
"loss": 0.0001,
"step": 16175
},
{
"epoch": 57.14,
"learning_rate": 1.9512820512820513e-07,
"loss": 0.0001,
"step": 16200
},
{
"epoch": 57.23,
"learning_rate": 1.9384615384615383e-07,
"loss": 0.0001,
"step": 16225
},
{
"epoch": 57.32,
"learning_rate": 1.9256410256410256e-07,
"loss": 0.0001,
"step": 16250
},
{
"epoch": 57.41,
"learning_rate": 1.9128205128205128e-07,
"loss": 0.0001,
"step": 16275
},
{
"epoch": 57.5,
"learning_rate": 1.8999999999999998e-07,
"loss": 0.0001,
"step": 16300
},
{
"epoch": 57.58,
"learning_rate": 1.887179487179487e-07,
"loss": 0.0001,
"step": 16325
},
{
"epoch": 57.67,
"learning_rate": 1.8743589743589743e-07,
"loss": 0.0001,
"step": 16350
},
{
"epoch": 57.76,
"learning_rate": 1.8615384615384613e-07,
"loss": 0.0001,
"step": 16375
},
{
"epoch": 57.85,
"learning_rate": 1.8487179487179485e-07,
"loss": 0.0001,
"step": 16400
},
{
"epoch": 57.94,
"learning_rate": 1.8358974358974358e-07,
"loss": 0.0001,
"step": 16425
},
{
"epoch": 58.02,
"learning_rate": 1.823076923076923e-07,
"loss": 0.0001,
"step": 16450
},
{
"epoch": 58.11,
"learning_rate": 1.8102564102564103e-07,
"loss": 0.0001,
"step": 16475
},
{
"epoch": 58.2,
"learning_rate": 1.7974358974358975e-07,
"loss": 0.0001,
"step": 16500
},
{
"epoch": 58.29,
"learning_rate": 1.7846153846153845e-07,
"loss": 0.0001,
"step": 16525
},
{
"epoch": 58.38,
"learning_rate": 1.7717948717948718e-07,
"loss": 0.0001,
"step": 16550
},
{
"epoch": 58.47,
"learning_rate": 1.7589743589743587e-07,
"loss": 0.0001,
"step": 16575
},
{
"epoch": 58.55,
"learning_rate": 1.746153846153846e-07,
"loss": 0.0001,
"step": 16600
},
{
"epoch": 58.64,
"learning_rate": 1.7333333333333332e-07,
"loss": 0.0001,
"step": 16625
},
{
"epoch": 58.73,
"learning_rate": 1.7205128205128205e-07,
"loss": 0.0001,
"step": 16650
},
{
"epoch": 58.82,
"learning_rate": 1.7076923076923075e-07,
"loss": 0.0001,
"step": 16675
},
{
"epoch": 58.91,
"learning_rate": 1.6948717948717947e-07,
"loss": 0.0001,
"step": 16700
},
{
"epoch": 58.99,
"learning_rate": 1.682051282051282e-07,
"loss": 0.0001,
"step": 16725
},
{
"epoch": 59.08,
"learning_rate": 1.6692307692307692e-07,
"loss": 0.0001,
"step": 16750
},
{
"epoch": 59.17,
"learning_rate": 1.6564102564102565e-07,
"loss": 0.0001,
"step": 16775
},
{
"epoch": 59.26,
"learning_rate": 1.6435897435897437e-07,
"loss": 0.0001,
"step": 16800
},
{
"epoch": 59.35,
"learning_rate": 1.6307692307692307e-07,
"loss": 0.0001,
"step": 16825
},
{
"epoch": 59.44,
"learning_rate": 1.6179487179487177e-07,
"loss": 0.0001,
"step": 16850
},
{
"epoch": 59.52,
"learning_rate": 1.605128205128205e-07,
"loss": 0.0001,
"step": 16875
},
{
"epoch": 59.61,
"learning_rate": 1.5923076923076922e-07,
"loss": 0.0001,
"step": 16900
},
{
"epoch": 59.7,
"learning_rate": 1.5794871794871794e-07,
"loss": 0.0001,
"step": 16925
},
{
"epoch": 59.79,
"learning_rate": 1.5666666666666667e-07,
"loss": 0.0001,
"step": 16950
},
{
"epoch": 59.88,
"learning_rate": 1.5538461538461536e-07,
"loss": 0.0001,
"step": 16975
},
{
"epoch": 59.96,
"learning_rate": 1.541025641025641e-07,
"loss": 0.0001,
"step": 17000
},
{
"epoch": 59.96,
"eval_loss": 0.42113855481147766,
"eval_runtime": 2319.7585,
"eval_samples_per_second": 3.911,
"eval_steps_per_second": 0.244,
"eval_wer": 4.58548286399632,
"step": 17000
},
{
"epoch": 60.05,
"learning_rate": 1.5282051282051281e-07,
"loss": 0.0001,
"step": 17025
},
{
"epoch": 60.14,
"learning_rate": 1.5153846153846154e-07,
"loss": 0.0001,
"step": 17050
},
{
"epoch": 60.23,
"learning_rate": 1.5025641025641026e-07,
"loss": 0.0001,
"step": 17075
},
{
"epoch": 60.32,
"learning_rate": 1.48974358974359e-07,
"loss": 0.0001,
"step": 17100
},
{
"epoch": 60.41,
"learning_rate": 1.476923076923077e-07,
"loss": 0.0001,
"step": 17125
},
{
"epoch": 60.49,
"learning_rate": 1.4641025641025639e-07,
"loss": 0.0001,
"step": 17150
},
{
"epoch": 60.58,
"learning_rate": 1.451282051282051e-07,
"loss": 0.0001,
"step": 17175
},
{
"epoch": 60.67,
"learning_rate": 1.4384615384615384e-07,
"loss": 0.0001,
"step": 17200
},
{
"epoch": 60.76,
"learning_rate": 1.4256410256410256e-07,
"loss": 0.0001,
"step": 17225
},
{
"epoch": 60.85,
"learning_rate": 1.4128205128205128e-07,
"loss": 0.0001,
"step": 17250
},
{
"epoch": 60.93,
"learning_rate": 1.4e-07,
"loss": 0.0001,
"step": 17275
},
{
"epoch": 61.02,
"learning_rate": 1.387179487179487e-07,
"loss": 0.0001,
"step": 17300
},
{
"epoch": 61.11,
"learning_rate": 1.3743589743589743e-07,
"loss": 0.0001,
"step": 17325
},
{
"epoch": 61.2,
"learning_rate": 1.3615384615384616e-07,
"loss": 0.0001,
"step": 17350
},
{
"epoch": 61.29,
"learning_rate": 1.3487179487179488e-07,
"loss": 0.0001,
"step": 17375
},
{
"epoch": 61.38,
"learning_rate": 1.3358974358974358e-07,
"loss": 0.0001,
"step": 17400
},
{
"epoch": 61.46,
"learning_rate": 1.323076923076923e-07,
"loss": 0.0001,
"step": 17425
},
{
"epoch": 61.55,
"learning_rate": 1.31025641025641e-07,
"loss": 0.0001,
"step": 17450
},
{
"epoch": 61.64,
"learning_rate": 1.2974358974358973e-07,
"loss": 0.0001,
"step": 17475
},
{
"epoch": 61.73,
"learning_rate": 1.2846153846153845e-07,
"loss": 0.0001,
"step": 17500
},
{
"epoch": 61.82,
"learning_rate": 1.2717948717948718e-07,
"loss": 0.0001,
"step": 17525
},
{
"epoch": 61.9,
"learning_rate": 1.258974358974359e-07,
"loss": 0.0001,
"step": 17550
},
{
"epoch": 61.99,
"learning_rate": 1.246153846153846e-07,
"loss": 0.0001,
"step": 17575
},
{
"epoch": 62.08,
"learning_rate": 1.2333333333333333e-07,
"loss": 0.0001,
"step": 17600
},
{
"epoch": 62.17,
"learning_rate": 1.2205128205128205e-07,
"loss": 0.0001,
"step": 17625
},
{
"epoch": 62.26,
"learning_rate": 1.2076923076923075e-07,
"loss": 0.0001,
"step": 17650
},
{
"epoch": 62.35,
"learning_rate": 1.1948717948717947e-07,
"loss": 0.0001,
"step": 17675
},
{
"epoch": 62.43,
"learning_rate": 1.182051282051282e-07,
"loss": 0.0001,
"step": 17700
},
{
"epoch": 62.52,
"learning_rate": 1.1692307692307691e-07,
"loss": 0.0001,
"step": 17725
},
{
"epoch": 62.61,
"learning_rate": 1.1564102564102564e-07,
"loss": 0.0001,
"step": 17750
},
{
"epoch": 62.7,
"learning_rate": 1.1435897435897435e-07,
"loss": 0.0001,
"step": 17775
},
{
"epoch": 62.79,
"learning_rate": 1.1307692307692307e-07,
"loss": 0.0001,
"step": 17800
},
{
"epoch": 62.87,
"learning_rate": 1.117948717948718e-07,
"loss": 0.0001,
"step": 17825
},
{
"epoch": 62.96,
"learning_rate": 1.105128205128205e-07,
"loss": 0.0001,
"step": 17850
},
{
"epoch": 63.05,
"learning_rate": 1.0923076923076922e-07,
"loss": 0.0001,
"step": 17875
},
{
"epoch": 63.14,
"learning_rate": 1.0794871794871794e-07,
"loss": 0.0001,
"step": 17900
},
{
"epoch": 63.23,
"learning_rate": 1.0666666666666667e-07,
"loss": 0.0001,
"step": 17925
},
{
"epoch": 63.32,
"learning_rate": 1.0538461538461538e-07,
"loss": 0.0001,
"step": 17950
},
{
"epoch": 63.4,
"learning_rate": 1.041025641025641e-07,
"loss": 0.0001,
"step": 17975
},
{
"epoch": 63.49,
"learning_rate": 1.0282051282051282e-07,
"loss": 0.0001,
"step": 18000
},
{
"epoch": 63.49,
"eval_loss": 0.42647331953048706,
"eval_runtime": 2297.8831,
"eval_samples_per_second": 3.948,
"eval_steps_per_second": 0.247,
"eval_wer": 4.583839910623336,
"step": 18000
},
{
"epoch": 63.58,
"learning_rate": 1.0153846153846153e-07,
"loss": 0.0001,
"step": 18025
},
{
"epoch": 63.67,
"learning_rate": 1.0025641025641025e-07,
"loss": 0.0001,
"step": 18050
},
{
"epoch": 63.76,
"learning_rate": 9.897435897435898e-08,
"loss": 0.0001,
"step": 18075
},
{
"epoch": 63.84,
"learning_rate": 9.769230769230769e-08,
"loss": 0.0001,
"step": 18100
},
{
"epoch": 63.93,
"learning_rate": 9.64102564102564e-08,
"loss": 0.0001,
"step": 18125
},
{
"epoch": 64.02,
"learning_rate": 9.512820512820513e-08,
"loss": 0.0001,
"step": 18150
},
{
"epoch": 64.11,
"learning_rate": 9.384615384615384e-08,
"loss": 0.0001,
"step": 18175
},
{
"epoch": 64.2,
"learning_rate": 9.256410256410256e-08,
"loss": 0.0001,
"step": 18200
},
{
"epoch": 64.29,
"learning_rate": 9.128205128205129e-08,
"loss": 0.0001,
"step": 18225
},
{
"epoch": 64.37,
"learning_rate": 9e-08,
"loss": 0.0001,
"step": 18250
},
{
"epoch": 64.46,
"learning_rate": 8.871794871794871e-08,
"loss": 0.0001,
"step": 18275
},
{
"epoch": 64.55,
"learning_rate": 8.743589743589743e-08,
"loss": 0.0001,
"step": 18300
},
{
"epoch": 64.64,
"learning_rate": 8.615384615384615e-08,
"loss": 0.0001,
"step": 18325
},
{
"epoch": 64.73,
"learning_rate": 8.487179487179487e-08,
"loss": 0.0001,
"step": 18350
},
{
"epoch": 64.81,
"learning_rate": 8.35897435897436e-08,
"loss": 0.0001,
"step": 18375
},
{
"epoch": 64.9,
"learning_rate": 8.23076923076923e-08,
"loss": 0.0001,
"step": 18400
},
{
"epoch": 64.99,
"learning_rate": 8.102564102564102e-08,
"loss": 0.0001,
"step": 18425
},
{
"epoch": 65.08,
"learning_rate": 7.974358974358974e-08,
"loss": 0.0001,
"step": 18450
},
{
"epoch": 65.17,
"learning_rate": 7.846153846153846e-08,
"loss": 0.0001,
"step": 18475
},
{
"epoch": 65.26,
"learning_rate": 7.717948717948718e-08,
"loss": 0.0001,
"step": 18500
},
{
"epoch": 65.34,
"learning_rate": 7.58974358974359e-08,
"loss": 0.0001,
"step": 18525
},
{
"epoch": 65.43,
"learning_rate": 7.46153846153846e-08,
"loss": 0.0001,
"step": 18550
},
{
"epoch": 65.52,
"learning_rate": 7.333333333333333e-08,
"loss": 0.0001,
"step": 18575
},
{
"epoch": 65.61,
"learning_rate": 7.205128205128205e-08,
"loss": 0.0001,
"step": 18600
},
{
"epoch": 65.7,
"learning_rate": 7.076923076923076e-08,
"loss": 0.0001,
"step": 18625
},
{
"epoch": 65.78,
"learning_rate": 6.948717948717949e-08,
"loss": 0.0001,
"step": 18650
},
{
"epoch": 65.87,
"learning_rate": 6.82051282051282e-08,
"loss": 0.0001,
"step": 18675
},
{
"epoch": 65.96,
"learning_rate": 6.692307692307691e-08,
"loss": 0.0001,
"step": 18700
},
{
"epoch": 66.05,
"learning_rate": 6.564102564102564e-08,
"loss": 0.0001,
"step": 18725
},
{
"epoch": 66.14,
"learning_rate": 6.435897435897436e-08,
"loss": 0.0001,
"step": 18750
},
{
"epoch": 66.23,
"learning_rate": 6.307692307692307e-08,
"loss": 0.0001,
"step": 18775
},
{
"epoch": 66.31,
"learning_rate": 6.179487179487179e-08,
"loss": 0.0001,
"step": 18800
},
{
"epoch": 66.4,
"learning_rate": 6.051282051282051e-08,
"loss": 0.0001,
"step": 18825
},
{
"epoch": 66.49,
"learning_rate": 5.923076923076923e-08,
"loss": 0.0001,
"step": 18850
},
{
"epoch": 66.58,
"learning_rate": 5.7948717948717946e-08,
"loss": 0.0001,
"step": 18875
},
{
"epoch": 66.67,
"learning_rate": 5.666666666666666e-08,
"loss": 0.0001,
"step": 18900
},
{
"epoch": 66.75,
"learning_rate": 5.538461538461538e-08,
"loss": 0.0001,
"step": 18925
},
{
"epoch": 66.84,
"learning_rate": 5.41025641025641e-08,
"loss": 0.0001,
"step": 18950
},
{
"epoch": 66.93,
"learning_rate": 5.282051282051281e-08,
"loss": 0.0001,
"step": 18975
},
{
"epoch": 67.02,
"learning_rate": 5.153846153846154e-08,
"loss": 0.0001,
"step": 19000
},
{
"epoch": 67.02,
"eval_loss": 0.4302147626876831,
"eval_runtime": 2396.1247,
"eval_samples_per_second": 3.786,
"eval_steps_per_second": 0.237,
"eval_wer": 4.592054677488252,
"step": 19000
},
{
"epoch": 67.11,
"learning_rate": 5.0256410256410255e-08,
"loss": 0.0001,
"step": 19025
},
{
"epoch": 67.2,
"learning_rate": 4.897435897435897e-08,
"loss": 0.0001,
"step": 19050
},
{
"epoch": 67.28,
"learning_rate": 4.769230769230769e-08,
"loss": 0.0001,
"step": 19075
},
{
"epoch": 67.37,
"learning_rate": 4.641025641025641e-08,
"loss": 0.0001,
"step": 19100
},
{
"epoch": 67.46,
"learning_rate": 4.512820512820513e-08,
"loss": 0.0001,
"step": 19125
},
{
"epoch": 67.55,
"learning_rate": 4.3846153846153846e-08,
"loss": 0.0001,
"step": 19150
},
{
"epoch": 67.64,
"learning_rate": 4.256410256410256e-08,
"loss": 0.0001,
"step": 19175
},
{
"epoch": 67.72,
"learning_rate": 4.128205128205128e-08,
"loss": 0.0001,
"step": 19200
},
{
"epoch": 67.81,
"learning_rate": 4e-08,
"loss": 0.0001,
"step": 19225
},
{
"epoch": 67.9,
"learning_rate": 3.871794871794871e-08,
"loss": 0.0001,
"step": 19250
},
{
"epoch": 67.99,
"learning_rate": 3.743589743589744e-08,
"loss": 0.0001,
"step": 19275
},
{
"epoch": 68.08,
"learning_rate": 3.6153846153846155e-08,
"loss": 0.0001,
"step": 19300
},
{
"epoch": 68.17,
"learning_rate": 3.487179487179487e-08,
"loss": 0.0001,
"step": 19325
},
{
"epoch": 68.25,
"learning_rate": 3.358974358974359e-08,
"loss": 0.0001,
"step": 19350
},
{
"epoch": 68.34,
"learning_rate": 3.230769230769231e-08,
"loss": 0.0001,
"step": 19375
},
{
"epoch": 68.43,
"learning_rate": 3.102564102564103e-08,
"loss": 0.0001,
"step": 19400
},
{
"epoch": 68.52,
"learning_rate": 2.9743589743589743e-08,
"loss": 0.0001,
"step": 19425
},
{
"epoch": 68.61,
"learning_rate": 2.846153846153846e-08,
"loss": 0.0001,
"step": 19450
},
{
"epoch": 68.69,
"learning_rate": 2.717948717948718e-08,
"loss": 0.0001,
"step": 19475
},
{
"epoch": 68.78,
"learning_rate": 2.5897435897435894e-08,
"loss": 0.0001,
"step": 19500
},
{
"epoch": 68.87,
"learning_rate": 2.4615384615384616e-08,
"loss": 0.0001,
"step": 19525
},
{
"epoch": 68.96,
"learning_rate": 2.3333333333333334e-08,
"loss": 0.0001,
"step": 19550
},
{
"epoch": 69.05,
"learning_rate": 2.205128205128205e-08,
"loss": 0.0001,
"step": 19575
},
{
"epoch": 69.14,
"learning_rate": 2.0769230769230767e-08,
"loss": 0.0001,
"step": 19600
},
{
"epoch": 69.22,
"learning_rate": 1.948717948717949e-08,
"loss": 0.0001,
"step": 19625
},
{
"epoch": 69.31,
"learning_rate": 1.8205128205128203e-08,
"loss": 0.0001,
"step": 19650
},
{
"epoch": 69.4,
"learning_rate": 1.692307692307692e-08,
"loss": 0.0001,
"step": 19675
},
{
"epoch": 69.49,
"learning_rate": 1.564102564102564e-08,
"loss": 0.0001,
"step": 19700
},
{
"epoch": 69.58,
"learning_rate": 1.4358974358974358e-08,
"loss": 0.0001,
"step": 19725
},
{
"epoch": 69.66,
"learning_rate": 1.3076923076923076e-08,
"loss": 0.0001,
"step": 19750
},
{
"epoch": 69.75,
"learning_rate": 1.1794871794871794e-08,
"loss": 0.0001,
"step": 19775
},
{
"epoch": 69.84,
"learning_rate": 1.0512820512820512e-08,
"loss": 0.0001,
"step": 19800
},
{
"epoch": 69.93,
"learning_rate": 9.23076923076923e-09,
"loss": 0.0001,
"step": 19825
},
{
"epoch": 70.02,
"learning_rate": 7.948717948717949e-09,
"loss": 0.0001,
"step": 19850
},
{
"epoch": 70.11,
"learning_rate": 6.666666666666667e-09,
"loss": 0.0001,
"step": 19875
},
{
"epoch": 70.19,
"learning_rate": 5.384615384615384e-09,
"loss": 0.0001,
"step": 19900
},
{
"epoch": 70.28,
"learning_rate": 4.102564102564102e-09,
"loss": 0.0001,
"step": 19925
},
{
"epoch": 70.37,
"learning_rate": 2.8205128205128205e-09,
"loss": 0.0001,
"step": 19950
},
{
"epoch": 70.46,
"learning_rate": 1.5384615384615385e-09,
"loss": 0.0001,
"step": 19975
},
{
"epoch": 70.55,
"learning_rate": 2.564102564102564e-10,
"loss": 0.0001,
"step": 20000
},
{
"epoch": 70.55,
"eval_loss": 0.4314596652984619,
"eval_runtime": 2359.2654,
"eval_samples_per_second": 3.845,
"eval_steps_per_second": 0.24,
"eval_wer": 4.600269444353169,
"step": 20000
},
{
"epoch": 70.55,
"step": 20000,
"total_flos": 4.3487978323968e+21,
"train_loss": 0.013711566552898149,
"train_runtime": 274306.7863,
"train_samples_per_second": 4.666,
"train_steps_per_second": 0.073
}
],
"logging_steps": 25,
"max_steps": 20000,
"num_input_tokens_seen": 0,
"num_train_epochs": 71,
"save_steps": 1000,
"total_flos": 4.3487978323968e+21,
"train_batch_size": 32,
"trial_name": null,
"trial_params": null
}