whisper-large-v2-eu-train / trainer_state.json
zuazo's picture
End of training
f994a5e verified
{
"best_metric": 12.512845312216648,
"best_model_checkpoint": "./checkpoint-19000",
"epoch": 116.95906432748538,
"eval_steps": 1000,
"global_step": 20000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.15,
"learning_rate": 5.000000000000001e-07,
"loss": 3.0272,
"step": 25
},
{
"epoch": 0.29,
"learning_rate": 1.0000000000000002e-06,
"loss": 2.1366,
"step": 50
},
{
"epoch": 0.44,
"learning_rate": 1.5e-06,
"loss": 1.43,
"step": 75
},
{
"epoch": 0.58,
"learning_rate": 2.0000000000000003e-06,
"loss": 1.164,
"step": 100
},
{
"epoch": 0.73,
"learning_rate": 2.5e-06,
"loss": 0.974,
"step": 125
},
{
"epoch": 0.88,
"learning_rate": 3e-06,
"loss": 0.828,
"step": 150
},
{
"epoch": 1.02,
"learning_rate": 3.5e-06,
"loss": 0.7451,
"step": 175
},
{
"epoch": 1.17,
"learning_rate": 4.000000000000001e-06,
"loss": 0.6279,
"step": 200
},
{
"epoch": 1.32,
"learning_rate": 4.5e-06,
"loss": 0.5869,
"step": 225
},
{
"epoch": 1.46,
"learning_rate": 5e-06,
"loss": 0.5309,
"step": 250
},
{
"epoch": 1.61,
"learning_rate": 5.500000000000001e-06,
"loss": 0.4996,
"step": 275
},
{
"epoch": 1.75,
"learning_rate": 6e-06,
"loss": 0.4623,
"step": 300
},
{
"epoch": 1.9,
"learning_rate": 6.5000000000000004e-06,
"loss": 0.4364,
"step": 325
},
{
"epoch": 2.05,
"learning_rate": 7e-06,
"loss": 0.3915,
"step": 350
},
{
"epoch": 2.19,
"learning_rate": 7.500000000000001e-06,
"loss": 0.3361,
"step": 375
},
{
"epoch": 2.34,
"learning_rate": 8.000000000000001e-06,
"loss": 0.3374,
"step": 400
},
{
"epoch": 2.49,
"learning_rate": 8.5e-06,
"loss": 0.3197,
"step": 425
},
{
"epoch": 2.63,
"learning_rate": 9e-06,
"loss": 0.3097,
"step": 450
},
{
"epoch": 2.78,
"learning_rate": 9.5e-06,
"loss": 0.2999,
"step": 475
},
{
"epoch": 2.92,
"learning_rate": 1e-05,
"loss": 0.2852,
"step": 500
},
{
"epoch": 3.07,
"learning_rate": 9.987179487179488e-06,
"loss": 0.2638,
"step": 525
},
{
"epoch": 3.22,
"learning_rate": 9.974358974358974e-06,
"loss": 0.2294,
"step": 550
},
{
"epoch": 3.36,
"learning_rate": 9.961538461538463e-06,
"loss": 0.2149,
"step": 575
},
{
"epoch": 3.51,
"learning_rate": 9.94871794871795e-06,
"loss": 0.2168,
"step": 600
},
{
"epoch": 3.65,
"learning_rate": 9.935897435897437e-06,
"loss": 0.2114,
"step": 625
},
{
"epoch": 3.8,
"learning_rate": 9.923076923076923e-06,
"loss": 0.2067,
"step": 650
},
{
"epoch": 3.95,
"learning_rate": 9.910256410256412e-06,
"loss": 0.1998,
"step": 675
},
{
"epoch": 4.09,
"learning_rate": 9.897435897435899e-06,
"loss": 0.1672,
"step": 700
},
{
"epoch": 4.24,
"learning_rate": 9.884615384615386e-06,
"loss": 0.1528,
"step": 725
},
{
"epoch": 4.39,
"learning_rate": 9.871794871794872e-06,
"loss": 0.1499,
"step": 750
},
{
"epoch": 4.53,
"learning_rate": 9.858974358974361e-06,
"loss": 0.1507,
"step": 775
},
{
"epoch": 4.68,
"learning_rate": 9.846153846153848e-06,
"loss": 0.1466,
"step": 800
},
{
"epoch": 4.82,
"learning_rate": 9.833333333333333e-06,
"loss": 0.1448,
"step": 825
},
{
"epoch": 4.97,
"learning_rate": 9.820512820512821e-06,
"loss": 0.1492,
"step": 850
},
{
"epoch": 5.12,
"learning_rate": 9.807692307692308e-06,
"loss": 0.1175,
"step": 875
},
{
"epoch": 5.26,
"learning_rate": 9.794871794871795e-06,
"loss": 0.1093,
"step": 900
},
{
"epoch": 5.41,
"learning_rate": 9.782051282051282e-06,
"loss": 0.1059,
"step": 925
},
{
"epoch": 5.56,
"learning_rate": 9.76923076923077e-06,
"loss": 0.1097,
"step": 950
},
{
"epoch": 5.7,
"learning_rate": 9.756410256410257e-06,
"loss": 0.108,
"step": 975
},
{
"epoch": 5.85,
"learning_rate": 9.743589743589744e-06,
"loss": 0.1098,
"step": 1000
},
{
"epoch": 5.85,
"eval_loss": 0.24953961372375488,
"eval_runtime": 1416.1968,
"eval_samples_per_second": 4.654,
"eval_steps_per_second": 0.291,
"eval_wer": 16.635434927159523,
"step": 1000
},
{
"epoch": 5.99,
"learning_rate": 9.730769230769231e-06,
"loss": 0.1101,
"step": 1025
},
{
"epoch": 6.14,
"learning_rate": 9.71794871794872e-06,
"loss": 0.0786,
"step": 1050
},
{
"epoch": 6.29,
"learning_rate": 9.705128205128206e-06,
"loss": 0.0742,
"step": 1075
},
{
"epoch": 6.43,
"learning_rate": 9.692307692307693e-06,
"loss": 0.0781,
"step": 1100
},
{
"epoch": 6.58,
"learning_rate": 9.67948717948718e-06,
"loss": 0.0781,
"step": 1125
},
{
"epoch": 6.73,
"learning_rate": 9.666666666666667e-06,
"loss": 0.0771,
"step": 1150
},
{
"epoch": 6.87,
"learning_rate": 9.653846153846155e-06,
"loss": 0.0794,
"step": 1175
},
{
"epoch": 7.02,
"learning_rate": 9.641025641025642e-06,
"loss": 0.0738,
"step": 1200
},
{
"epoch": 7.16,
"learning_rate": 9.628205128205129e-06,
"loss": 0.0557,
"step": 1225
},
{
"epoch": 7.31,
"learning_rate": 9.615384615384616e-06,
"loss": 0.0561,
"step": 1250
},
{
"epoch": 7.46,
"learning_rate": 9.602564102564104e-06,
"loss": 0.0576,
"step": 1275
},
{
"epoch": 7.6,
"learning_rate": 9.589743589743591e-06,
"loss": 0.0577,
"step": 1300
},
{
"epoch": 7.75,
"learning_rate": 9.576923076923078e-06,
"loss": 0.0601,
"step": 1325
},
{
"epoch": 7.89,
"learning_rate": 9.564102564102565e-06,
"loss": 0.0599,
"step": 1350
},
{
"epoch": 8.04,
"learning_rate": 9.551282051282053e-06,
"loss": 0.0543,
"step": 1375
},
{
"epoch": 8.19,
"learning_rate": 9.53846153846154e-06,
"loss": 0.0432,
"step": 1400
},
{
"epoch": 8.33,
"learning_rate": 9.525641025641025e-06,
"loss": 0.0405,
"step": 1425
},
{
"epoch": 8.48,
"learning_rate": 9.512820512820514e-06,
"loss": 0.0427,
"step": 1450
},
{
"epoch": 8.63,
"learning_rate": 9.5e-06,
"loss": 0.0445,
"step": 1475
},
{
"epoch": 8.77,
"learning_rate": 9.487179487179487e-06,
"loss": 0.0451,
"step": 1500
},
{
"epoch": 8.92,
"learning_rate": 9.474358974358974e-06,
"loss": 0.0479,
"step": 1525
},
{
"epoch": 9.06,
"learning_rate": 9.461538461538463e-06,
"loss": 0.0407,
"step": 1550
},
{
"epoch": 9.21,
"learning_rate": 9.44871794871795e-06,
"loss": 0.0328,
"step": 1575
},
{
"epoch": 9.36,
"learning_rate": 9.435897435897436e-06,
"loss": 0.034,
"step": 1600
},
{
"epoch": 9.5,
"learning_rate": 9.423076923076923e-06,
"loss": 0.0343,
"step": 1625
},
{
"epoch": 9.65,
"learning_rate": 9.410256410256412e-06,
"loss": 0.0334,
"step": 1650
},
{
"epoch": 9.8,
"learning_rate": 9.397435897435899e-06,
"loss": 0.0339,
"step": 1675
},
{
"epoch": 9.94,
"learning_rate": 9.384615384615385e-06,
"loss": 0.0339,
"step": 1700
},
{
"epoch": 10.09,
"learning_rate": 9.371794871794872e-06,
"loss": 0.0294,
"step": 1725
},
{
"epoch": 10.23,
"learning_rate": 9.358974358974359e-06,
"loss": 0.0232,
"step": 1750
},
{
"epoch": 10.38,
"learning_rate": 9.346153846153847e-06,
"loss": 0.0236,
"step": 1775
},
{
"epoch": 10.53,
"learning_rate": 9.333333333333334e-06,
"loss": 0.025,
"step": 1800
},
{
"epoch": 10.67,
"learning_rate": 9.320512820512821e-06,
"loss": 0.0284,
"step": 1825
},
{
"epoch": 10.82,
"learning_rate": 9.307692307692308e-06,
"loss": 0.0268,
"step": 1850
},
{
"epoch": 10.96,
"learning_rate": 9.294871794871796e-06,
"loss": 0.0262,
"step": 1875
},
{
"epoch": 11.11,
"learning_rate": 9.282051282051283e-06,
"loss": 0.0209,
"step": 1900
},
{
"epoch": 11.26,
"learning_rate": 9.26923076923077e-06,
"loss": 0.0192,
"step": 1925
},
{
"epoch": 11.4,
"learning_rate": 9.256410256410257e-06,
"loss": 0.0191,
"step": 1950
},
{
"epoch": 11.55,
"learning_rate": 9.243589743589745e-06,
"loss": 0.0201,
"step": 1975
},
{
"epoch": 11.7,
"learning_rate": 9.230769230769232e-06,
"loss": 0.022,
"step": 2000
},
{
"epoch": 11.7,
"eval_loss": 0.2732812166213989,
"eval_runtime": 1412.7854,
"eval_samples_per_second": 4.665,
"eval_steps_per_second": 0.292,
"eval_wer": 14.630558745894536,
"step": 2000
},
{
"epoch": 11.84,
"learning_rate": 9.217948717948717e-06,
"loss": 0.0209,
"step": 2025
},
{
"epoch": 11.99,
"learning_rate": 9.205128205128206e-06,
"loss": 0.021,
"step": 2050
},
{
"epoch": 12.13,
"learning_rate": 9.192307692307693e-06,
"loss": 0.0182,
"step": 2075
},
{
"epoch": 12.28,
"learning_rate": 9.17948717948718e-06,
"loss": 0.017,
"step": 2100
},
{
"epoch": 12.43,
"learning_rate": 9.166666666666666e-06,
"loss": 0.0168,
"step": 2125
},
{
"epoch": 12.57,
"learning_rate": 9.153846153846155e-06,
"loss": 0.0171,
"step": 2150
},
{
"epoch": 12.72,
"learning_rate": 9.141025641025642e-06,
"loss": 0.0168,
"step": 2175
},
{
"epoch": 12.87,
"learning_rate": 9.128205128205129e-06,
"loss": 0.0174,
"step": 2200
},
{
"epoch": 13.01,
"learning_rate": 9.115384615384615e-06,
"loss": 0.0168,
"step": 2225
},
{
"epoch": 13.16,
"learning_rate": 9.102564102564104e-06,
"loss": 0.0137,
"step": 2250
},
{
"epoch": 13.3,
"learning_rate": 9.08974358974359e-06,
"loss": 0.0135,
"step": 2275
},
{
"epoch": 13.45,
"learning_rate": 9.076923076923078e-06,
"loss": 0.0152,
"step": 2300
},
{
"epoch": 13.6,
"learning_rate": 9.064102564102564e-06,
"loss": 0.0143,
"step": 2325
},
{
"epoch": 13.74,
"learning_rate": 9.051282051282051e-06,
"loss": 0.014,
"step": 2350
},
{
"epoch": 13.89,
"learning_rate": 9.03846153846154e-06,
"loss": 0.0137,
"step": 2375
},
{
"epoch": 14.04,
"learning_rate": 9.025641025641027e-06,
"loss": 0.013,
"step": 2400
},
{
"epoch": 14.18,
"learning_rate": 9.012820512820513e-06,
"loss": 0.012,
"step": 2425
},
{
"epoch": 14.33,
"learning_rate": 9e-06,
"loss": 0.0121,
"step": 2450
},
{
"epoch": 14.47,
"learning_rate": 8.987179487179489e-06,
"loss": 0.0109,
"step": 2475
},
{
"epoch": 14.62,
"learning_rate": 8.974358974358976e-06,
"loss": 0.0132,
"step": 2500
},
{
"epoch": 14.77,
"learning_rate": 8.961538461538462e-06,
"loss": 0.0125,
"step": 2525
},
{
"epoch": 14.91,
"learning_rate": 8.94871794871795e-06,
"loss": 0.0126,
"step": 2550
},
{
"epoch": 15.06,
"learning_rate": 8.935897435897438e-06,
"loss": 0.0113,
"step": 2575
},
{
"epoch": 15.2,
"learning_rate": 8.923076923076925e-06,
"loss": 0.0108,
"step": 2600
},
{
"epoch": 15.35,
"learning_rate": 8.910256410256411e-06,
"loss": 0.0108,
"step": 2625
},
{
"epoch": 15.5,
"learning_rate": 8.897435897435898e-06,
"loss": 0.0095,
"step": 2650
},
{
"epoch": 15.64,
"learning_rate": 8.884615384615385e-06,
"loss": 0.0115,
"step": 2675
},
{
"epoch": 15.79,
"learning_rate": 8.871794871794872e-06,
"loss": 0.011,
"step": 2700
},
{
"epoch": 15.94,
"learning_rate": 8.858974358974359e-06,
"loss": 0.0117,
"step": 2725
},
{
"epoch": 16.08,
"learning_rate": 8.846153846153847e-06,
"loss": 0.0118,
"step": 2750
},
{
"epoch": 16.23,
"learning_rate": 8.833333333333334e-06,
"loss": 0.0094,
"step": 2775
},
{
"epoch": 16.37,
"learning_rate": 8.820512820512821e-06,
"loss": 0.0092,
"step": 2800
},
{
"epoch": 16.52,
"learning_rate": 8.807692307692308e-06,
"loss": 0.0099,
"step": 2825
},
{
"epoch": 16.67,
"learning_rate": 8.794871794871796e-06,
"loss": 0.0105,
"step": 2850
},
{
"epoch": 16.81,
"learning_rate": 8.782051282051283e-06,
"loss": 0.0094,
"step": 2875
},
{
"epoch": 16.96,
"learning_rate": 8.76923076923077e-06,
"loss": 0.0099,
"step": 2900
},
{
"epoch": 17.11,
"learning_rate": 8.756410256410257e-06,
"loss": 0.01,
"step": 2925
},
{
"epoch": 17.25,
"learning_rate": 8.743589743589743e-06,
"loss": 0.0088,
"step": 2950
},
{
"epoch": 17.4,
"learning_rate": 8.730769230769232e-06,
"loss": 0.0091,
"step": 2975
},
{
"epoch": 17.54,
"learning_rate": 8.717948717948719e-06,
"loss": 0.0089,
"step": 3000
},
{
"epoch": 17.54,
"eval_loss": 0.3074812889099121,
"eval_runtime": 1410.9888,
"eval_samples_per_second": 4.671,
"eval_steps_per_second": 0.292,
"eval_wer": 13.969654838904672,
"step": 3000
},
{
"epoch": 17.69,
"learning_rate": 8.705128205128206e-06,
"loss": 0.0094,
"step": 3025
},
{
"epoch": 17.84,
"learning_rate": 8.692307692307692e-06,
"loss": 0.0094,
"step": 3050
},
{
"epoch": 17.98,
"learning_rate": 8.679487179487181e-06,
"loss": 0.0095,
"step": 3075
},
{
"epoch": 18.13,
"learning_rate": 8.667179487179488e-06,
"loss": 0.008,
"step": 3100
},
{
"epoch": 18.27,
"learning_rate": 8.654358974358975e-06,
"loss": 0.0073,
"step": 3125
},
{
"epoch": 18.42,
"learning_rate": 8.641538461538462e-06,
"loss": 0.0084,
"step": 3150
},
{
"epoch": 18.57,
"learning_rate": 8.628717948717949e-06,
"loss": 0.008,
"step": 3175
},
{
"epoch": 18.71,
"learning_rate": 8.615897435897437e-06,
"loss": 0.0075,
"step": 3200
},
{
"epoch": 18.86,
"learning_rate": 8.603076923076924e-06,
"loss": 0.0081,
"step": 3225
},
{
"epoch": 19.01,
"learning_rate": 8.590256410256411e-06,
"loss": 0.009,
"step": 3250
},
{
"epoch": 19.15,
"learning_rate": 8.577435897435898e-06,
"loss": 0.0068,
"step": 3275
},
{
"epoch": 19.3,
"learning_rate": 8.564615384615386e-06,
"loss": 0.0076,
"step": 3300
},
{
"epoch": 19.44,
"learning_rate": 8.551794871794873e-06,
"loss": 0.0076,
"step": 3325
},
{
"epoch": 19.59,
"learning_rate": 8.53897435897436e-06,
"loss": 0.0078,
"step": 3350
},
{
"epoch": 19.74,
"learning_rate": 8.526153846153847e-06,
"loss": 0.0081,
"step": 3375
},
{
"epoch": 19.88,
"learning_rate": 8.513333333333335e-06,
"loss": 0.0085,
"step": 3400
},
{
"epoch": 20.03,
"learning_rate": 8.50051282051282e-06,
"loss": 0.0091,
"step": 3425
},
{
"epoch": 20.18,
"learning_rate": 8.487692307692307e-06,
"loss": 0.0073,
"step": 3450
},
{
"epoch": 20.32,
"learning_rate": 8.474871794871796e-06,
"loss": 0.0075,
"step": 3475
},
{
"epoch": 20.47,
"learning_rate": 8.462051282051283e-06,
"loss": 0.0076,
"step": 3500
},
{
"epoch": 20.61,
"learning_rate": 8.44923076923077e-06,
"loss": 0.0095,
"step": 3525
},
{
"epoch": 20.76,
"learning_rate": 8.436410256410256e-06,
"loss": 0.0092,
"step": 3550
},
{
"epoch": 20.91,
"learning_rate": 8.423589743589745e-06,
"loss": 0.0076,
"step": 3575
},
{
"epoch": 21.05,
"learning_rate": 8.410769230769232e-06,
"loss": 0.0076,
"step": 3600
},
{
"epoch": 21.2,
"learning_rate": 8.397948717948719e-06,
"loss": 0.0056,
"step": 3625
},
{
"epoch": 21.35,
"learning_rate": 8.385128205128205e-06,
"loss": 0.006,
"step": 3650
},
{
"epoch": 21.49,
"learning_rate": 8.372307692307692e-06,
"loss": 0.0061,
"step": 3675
},
{
"epoch": 21.64,
"learning_rate": 8.35948717948718e-06,
"loss": 0.0067,
"step": 3700
},
{
"epoch": 21.78,
"learning_rate": 8.346666666666668e-06,
"loss": 0.008,
"step": 3725
},
{
"epoch": 21.93,
"learning_rate": 8.333846153846154e-06,
"loss": 0.0075,
"step": 3750
},
{
"epoch": 22.08,
"learning_rate": 8.321025641025641e-06,
"loss": 0.0065,
"step": 3775
},
{
"epoch": 22.22,
"learning_rate": 8.30820512820513e-06,
"loss": 0.0065,
"step": 3800
},
{
"epoch": 22.37,
"learning_rate": 8.295384615384616e-06,
"loss": 0.0058,
"step": 3825
},
{
"epoch": 22.51,
"learning_rate": 8.282564102564103e-06,
"loss": 0.0062,
"step": 3850
},
{
"epoch": 22.66,
"learning_rate": 8.26974358974359e-06,
"loss": 0.0064,
"step": 3875
},
{
"epoch": 22.81,
"learning_rate": 8.256923076923079e-06,
"loss": 0.0069,
"step": 3900
},
{
"epoch": 22.95,
"learning_rate": 8.244102564102565e-06,
"loss": 0.0067,
"step": 3925
},
{
"epoch": 23.1,
"learning_rate": 8.231282051282052e-06,
"loss": 0.0056,
"step": 3950
},
{
"epoch": 23.25,
"learning_rate": 8.218461538461539e-06,
"loss": 0.005,
"step": 3975
},
{
"epoch": 23.39,
"learning_rate": 8.205641025641028e-06,
"loss": 0.0056,
"step": 4000
},
{
"epoch": 23.39,
"eval_loss": 0.3205597400665283,
"eval_runtime": 1412.0433,
"eval_samples_per_second": 4.668,
"eval_steps_per_second": 0.292,
"eval_wer": 14.072417336637852,
"step": 4000
},
{
"epoch": 23.54,
"learning_rate": 8.192820512820513e-06,
"loss": 0.005,
"step": 4025
},
{
"epoch": 23.68,
"learning_rate": 8.18e-06,
"loss": 0.0054,
"step": 4050
},
{
"epoch": 23.83,
"learning_rate": 8.167179487179488e-06,
"loss": 0.0067,
"step": 4075
},
{
"epoch": 23.98,
"learning_rate": 8.154358974358975e-06,
"loss": 0.0065,
"step": 4100
},
{
"epoch": 24.12,
"learning_rate": 8.141538461538462e-06,
"loss": 0.0049,
"step": 4125
},
{
"epoch": 24.27,
"learning_rate": 8.128717948717949e-06,
"loss": 0.0054,
"step": 4150
},
{
"epoch": 24.42,
"learning_rate": 8.115897435897437e-06,
"loss": 0.0047,
"step": 4175
},
{
"epoch": 24.56,
"learning_rate": 8.103076923076924e-06,
"loss": 0.0046,
"step": 4200
},
{
"epoch": 24.71,
"learning_rate": 8.09025641025641e-06,
"loss": 0.0045,
"step": 4225
},
{
"epoch": 24.85,
"learning_rate": 8.077435897435898e-06,
"loss": 0.0039,
"step": 4250
},
{
"epoch": 25.0,
"learning_rate": 8.064615384615384e-06,
"loss": 0.0052,
"step": 4275
},
{
"epoch": 25.15,
"learning_rate": 8.051794871794873e-06,
"loss": 0.0048,
"step": 4300
},
{
"epoch": 25.29,
"learning_rate": 8.03897435897436e-06,
"loss": 0.0047,
"step": 4325
},
{
"epoch": 25.44,
"learning_rate": 8.026153846153847e-06,
"loss": 0.0046,
"step": 4350
},
{
"epoch": 25.58,
"learning_rate": 8.013333333333333e-06,
"loss": 0.0044,
"step": 4375
},
{
"epoch": 25.73,
"learning_rate": 8.000512820512822e-06,
"loss": 0.0042,
"step": 4400
},
{
"epoch": 25.88,
"learning_rate": 7.987692307692309e-06,
"loss": 0.0053,
"step": 4425
},
{
"epoch": 26.02,
"learning_rate": 7.974871794871796e-06,
"loss": 0.0044,
"step": 4450
},
{
"epoch": 26.17,
"learning_rate": 7.962051282051282e-06,
"loss": 0.0032,
"step": 4475
},
{
"epoch": 26.32,
"learning_rate": 7.949230769230771e-06,
"loss": 0.0044,
"step": 4500
},
{
"epoch": 26.46,
"learning_rate": 7.936410256410258e-06,
"loss": 0.0061,
"step": 4525
},
{
"epoch": 26.61,
"learning_rate": 7.923589743589745e-06,
"loss": 0.0042,
"step": 4550
},
{
"epoch": 26.75,
"learning_rate": 7.910769230769231e-06,
"loss": 0.0048,
"step": 4575
},
{
"epoch": 26.9,
"learning_rate": 7.89794871794872e-06,
"loss": 0.0046,
"step": 4600
},
{
"epoch": 27.05,
"learning_rate": 7.885128205128207e-06,
"loss": 0.0049,
"step": 4625
},
{
"epoch": 27.19,
"learning_rate": 7.872307692307692e-06,
"loss": 0.0055,
"step": 4650
},
{
"epoch": 27.34,
"learning_rate": 7.85948717948718e-06,
"loss": 0.0039,
"step": 4675
},
{
"epoch": 27.49,
"learning_rate": 7.846666666666667e-06,
"loss": 0.0043,
"step": 4700
},
{
"epoch": 27.63,
"learning_rate": 7.833846153846154e-06,
"loss": 0.0045,
"step": 4725
},
{
"epoch": 27.78,
"learning_rate": 7.821025641025641e-06,
"loss": 0.0042,
"step": 4750
},
{
"epoch": 27.92,
"learning_rate": 7.80820512820513e-06,
"loss": 0.0052,
"step": 4775
},
{
"epoch": 28.07,
"learning_rate": 7.795384615384616e-06,
"loss": 0.0035,
"step": 4800
},
{
"epoch": 28.22,
"learning_rate": 7.782564102564103e-06,
"loss": 0.0038,
"step": 4825
},
{
"epoch": 28.36,
"learning_rate": 7.76974358974359e-06,
"loss": 0.0044,
"step": 4850
},
{
"epoch": 28.51,
"learning_rate": 7.756923076923077e-06,
"loss": 0.004,
"step": 4875
},
{
"epoch": 28.65,
"learning_rate": 7.744102564102565e-06,
"loss": 0.005,
"step": 4900
},
{
"epoch": 28.8,
"learning_rate": 7.731282051282052e-06,
"loss": 0.0051,
"step": 4925
},
{
"epoch": 28.95,
"learning_rate": 7.718461538461539e-06,
"loss": 0.005,
"step": 4950
},
{
"epoch": 29.09,
"learning_rate": 7.705641025641026e-06,
"loss": 0.0052,
"step": 4975
},
{
"epoch": 29.24,
"learning_rate": 7.692820512820514e-06,
"loss": 0.0053,
"step": 5000
},
{
"epoch": 29.24,
"eval_loss": 0.33138585090637207,
"eval_runtime": 1411.4202,
"eval_samples_per_second": 4.67,
"eval_steps_per_second": 0.292,
"eval_wer": 13.794354107477483,
"step": 5000
},
{
"epoch": 29.39,
"learning_rate": 7.680000000000001e-06,
"loss": 0.005,
"step": 5025
},
{
"epoch": 29.53,
"learning_rate": 7.667179487179488e-06,
"loss": 0.0056,
"step": 5050
},
{
"epoch": 29.68,
"learning_rate": 7.654358974358975e-06,
"loss": 0.0053,
"step": 5075
},
{
"epoch": 29.82,
"learning_rate": 7.641538461538463e-06,
"loss": 0.0048,
"step": 5100
},
{
"epoch": 29.97,
"learning_rate": 7.62871794871795e-06,
"loss": 0.005,
"step": 5125
},
{
"epoch": 30.12,
"learning_rate": 7.615897435897437e-06,
"loss": 0.0042,
"step": 5150
},
{
"epoch": 30.26,
"learning_rate": 7.6030769230769245e-06,
"loss": 0.0036,
"step": 5175
},
{
"epoch": 30.41,
"learning_rate": 7.590256410256411e-06,
"loss": 0.0034,
"step": 5200
},
{
"epoch": 30.56,
"learning_rate": 7.577435897435899e-06,
"loss": 0.0041,
"step": 5225
},
{
"epoch": 30.7,
"learning_rate": 7.564615384615385e-06,
"loss": 0.0045,
"step": 5250
},
{
"epoch": 30.85,
"learning_rate": 7.551794871794872e-06,
"loss": 0.0037,
"step": 5275
},
{
"epoch": 30.99,
"learning_rate": 7.5389743589743595e-06,
"loss": 0.0037,
"step": 5300
},
{
"epoch": 31.14,
"learning_rate": 7.526153846153846e-06,
"loss": 0.0038,
"step": 5325
},
{
"epoch": 31.29,
"learning_rate": 7.513333333333334e-06,
"loss": 0.0035,
"step": 5350
},
{
"epoch": 31.43,
"learning_rate": 7.500512820512821e-06,
"loss": 0.0032,
"step": 5375
},
{
"epoch": 31.58,
"learning_rate": 7.4876923076923085e-06,
"loss": 0.0041,
"step": 5400
},
{
"epoch": 31.73,
"learning_rate": 7.474871794871795e-06,
"loss": 0.0043,
"step": 5425
},
{
"epoch": 31.87,
"learning_rate": 7.462051282051282e-06,
"loss": 0.0044,
"step": 5450
},
{
"epoch": 32.02,
"learning_rate": 7.44923076923077e-06,
"loss": 0.0038,
"step": 5475
},
{
"epoch": 32.16,
"learning_rate": 7.436410256410257e-06,
"loss": 0.0037,
"step": 5500
},
{
"epoch": 32.31,
"learning_rate": 7.423589743589744e-06,
"loss": 0.0036,
"step": 5525
},
{
"epoch": 32.46,
"learning_rate": 7.410769230769231e-06,
"loss": 0.0034,
"step": 5550
},
{
"epoch": 32.6,
"learning_rate": 7.397948717948719e-06,
"loss": 0.0034,
"step": 5575
},
{
"epoch": 32.75,
"learning_rate": 7.385128205128206e-06,
"loss": 0.004,
"step": 5600
},
{
"epoch": 32.89,
"learning_rate": 7.372307692307693e-06,
"loss": 0.0038,
"step": 5625
},
{
"epoch": 33.04,
"learning_rate": 7.35948717948718e-06,
"loss": 0.0035,
"step": 5650
},
{
"epoch": 33.19,
"learning_rate": 7.346666666666668e-06,
"loss": 0.0033,
"step": 5675
},
{
"epoch": 33.33,
"learning_rate": 7.333846153846155e-06,
"loss": 0.0038,
"step": 5700
},
{
"epoch": 33.48,
"learning_rate": 7.321025641025642e-06,
"loss": 0.0028,
"step": 5725
},
{
"epoch": 33.63,
"learning_rate": 7.308205128205129e-06,
"loss": 0.0032,
"step": 5750
},
{
"epoch": 33.77,
"learning_rate": 7.295384615384617e-06,
"loss": 0.0041,
"step": 5775
},
{
"epoch": 33.92,
"learning_rate": 7.2825641025641036e-06,
"loss": 0.0043,
"step": 5800
},
{
"epoch": 34.06,
"learning_rate": 7.269743589743591e-06,
"loss": 0.004,
"step": 5825
},
{
"epoch": 34.21,
"learning_rate": 7.256923076923078e-06,
"loss": 0.0032,
"step": 5850
},
{
"epoch": 34.36,
"learning_rate": 7.244102564102564e-06,
"loss": 0.0029,
"step": 5875
},
{
"epoch": 34.5,
"learning_rate": 7.231282051282052e-06,
"loss": 0.0026,
"step": 5900
},
{
"epoch": 34.65,
"learning_rate": 7.2184615384615385e-06,
"loss": 0.0032,
"step": 5925
},
{
"epoch": 34.8,
"learning_rate": 7.205641025641026e-06,
"loss": 0.0033,
"step": 5950
},
{
"epoch": 34.94,
"learning_rate": 7.192820512820513e-06,
"loss": 0.0038,
"step": 5975
},
{
"epoch": 35.09,
"learning_rate": 7.180000000000001e-06,
"loss": 0.0037,
"step": 6000
},
{
"epoch": 35.09,
"eval_loss": 0.3375991880893707,
"eval_runtime": 1412.7642,
"eval_samples_per_second": 4.665,
"eval_steps_per_second": 0.292,
"eval_wer": 13.748010235950755,
"step": 6000
},
{
"epoch": 35.23,
"learning_rate": 7.1671794871794875e-06,
"loss": 0.0038,
"step": 6025
},
{
"epoch": 35.38,
"learning_rate": 7.154358974358974e-06,
"loss": 0.003,
"step": 6050
},
{
"epoch": 35.53,
"learning_rate": 7.141538461538462e-06,
"loss": 0.004,
"step": 6075
},
{
"epoch": 35.67,
"learning_rate": 7.128717948717949e-06,
"loss": 0.0041,
"step": 6100
},
{
"epoch": 35.82,
"learning_rate": 7.1158974358974365e-06,
"loss": 0.005,
"step": 6125
},
{
"epoch": 35.96,
"learning_rate": 7.103076923076923e-06,
"loss": 0.0053,
"step": 6150
},
{
"epoch": 36.11,
"learning_rate": 7.090256410256411e-06,
"loss": 0.0046,
"step": 6175
},
{
"epoch": 36.26,
"learning_rate": 7.077435897435898e-06,
"loss": 0.0033,
"step": 6200
},
{
"epoch": 36.4,
"learning_rate": 7.0646153846153855e-06,
"loss": 0.0032,
"step": 6225
},
{
"epoch": 36.55,
"learning_rate": 7.051794871794872e-06,
"loss": 0.0023,
"step": 6250
},
{
"epoch": 36.7,
"learning_rate": 7.03897435897436e-06,
"loss": 0.0028,
"step": 6275
},
{
"epoch": 36.84,
"learning_rate": 7.026153846153847e-06,
"loss": 0.0032,
"step": 6300
},
{
"epoch": 36.99,
"learning_rate": 7.0133333333333345e-06,
"loss": 0.0035,
"step": 6325
},
{
"epoch": 37.13,
"learning_rate": 7.000512820512821e-06,
"loss": 0.004,
"step": 6350
},
{
"epoch": 37.28,
"learning_rate": 6.987692307692309e-06,
"loss": 0.0032,
"step": 6375
},
{
"epoch": 37.43,
"learning_rate": 6.974871794871796e-06,
"loss": 0.0037,
"step": 6400
},
{
"epoch": 37.57,
"learning_rate": 6.9620512820512835e-06,
"loss": 0.0028,
"step": 6425
},
{
"epoch": 37.72,
"learning_rate": 6.94923076923077e-06,
"loss": 0.003,
"step": 6450
},
{
"epoch": 37.87,
"learning_rate": 6.936410256410256e-06,
"loss": 0.0029,
"step": 6475
},
{
"epoch": 38.01,
"learning_rate": 6.923589743589744e-06,
"loss": 0.0025,
"step": 6500
},
{
"epoch": 38.16,
"learning_rate": 6.910769230769231e-06,
"loss": 0.0027,
"step": 6525
},
{
"epoch": 38.3,
"learning_rate": 6.8979487179487185e-06,
"loss": 0.0021,
"step": 6550
},
{
"epoch": 38.45,
"learning_rate": 6.885128205128205e-06,
"loss": 0.0021,
"step": 6575
},
{
"epoch": 38.6,
"learning_rate": 6.872307692307693e-06,
"loss": 0.0025,
"step": 6600
},
{
"epoch": 38.74,
"learning_rate": 6.85948717948718e-06,
"loss": 0.0023,
"step": 6625
},
{
"epoch": 38.89,
"learning_rate": 6.846666666666667e-06,
"loss": 0.003,
"step": 6650
},
{
"epoch": 39.04,
"learning_rate": 6.833846153846154e-06,
"loss": 0.0032,
"step": 6675
},
{
"epoch": 39.18,
"learning_rate": 6.821025641025641e-06,
"loss": 0.0024,
"step": 6700
},
{
"epoch": 39.33,
"learning_rate": 6.808205128205129e-06,
"loss": 0.0022,
"step": 6725
},
{
"epoch": 39.47,
"learning_rate": 6.795384615384616e-06,
"loss": 0.0023,
"step": 6750
},
{
"epoch": 39.62,
"learning_rate": 6.782564102564103e-06,
"loss": 0.0019,
"step": 6775
},
{
"epoch": 39.77,
"learning_rate": 6.76974358974359e-06,
"loss": 0.0024,
"step": 6800
},
{
"epoch": 39.91,
"learning_rate": 6.756923076923078e-06,
"loss": 0.0024,
"step": 6825
},
{
"epoch": 40.06,
"learning_rate": 6.744102564102565e-06,
"loss": 0.0025,
"step": 6850
},
{
"epoch": 40.2,
"learning_rate": 6.731282051282052e-06,
"loss": 0.0024,
"step": 6875
},
{
"epoch": 40.35,
"learning_rate": 6.718461538461539e-06,
"loss": 0.0019,
"step": 6900
},
{
"epoch": 40.5,
"learning_rate": 6.705641025641027e-06,
"loss": 0.0017,
"step": 6925
},
{
"epoch": 40.64,
"learning_rate": 6.692820512820514e-06,
"loss": 0.0021,
"step": 6950
},
{
"epoch": 40.79,
"learning_rate": 6.680000000000001e-06,
"loss": 0.0029,
"step": 6975
},
{
"epoch": 40.94,
"learning_rate": 6.667179487179488e-06,
"loss": 0.0027,
"step": 7000
},
{
"epoch": 40.94,
"eval_loss": 0.34923917055130005,
"eval_runtime": 1411.1706,
"eval_samples_per_second": 4.671,
"eval_steps_per_second": 0.292,
"eval_wer": 13.68151685506458,
"step": 7000
},
{
"epoch": 41.08,
"learning_rate": 6.654358974358976e-06,
"loss": 0.0025,
"step": 7025
},
{
"epoch": 41.23,
"learning_rate": 6.641538461538463e-06,
"loss": 0.0022,
"step": 7050
},
{
"epoch": 41.37,
"learning_rate": 6.6287179487179486e-06,
"loss": 0.0019,
"step": 7075
},
{
"epoch": 41.52,
"learning_rate": 6.615897435897436e-06,
"loss": 0.002,
"step": 7100
},
{
"epoch": 41.67,
"learning_rate": 6.603076923076923e-06,
"loss": 0.0026,
"step": 7125
},
{
"epoch": 41.81,
"learning_rate": 6.590256410256411e-06,
"loss": 0.0029,
"step": 7150
},
{
"epoch": 41.96,
"learning_rate": 6.5774358974358976e-06,
"loss": 0.003,
"step": 7175
},
{
"epoch": 42.11,
"learning_rate": 6.564615384615385e-06,
"loss": 0.0029,
"step": 7200
},
{
"epoch": 42.25,
"learning_rate": 6.551794871794872e-06,
"loss": 0.0018,
"step": 7225
},
{
"epoch": 42.4,
"learning_rate": 6.538974358974359e-06,
"loss": 0.0019,
"step": 7250
},
{
"epoch": 42.54,
"learning_rate": 6.5261538461538465e-06,
"loss": 0.0025,
"step": 7275
},
{
"epoch": 42.69,
"learning_rate": 6.513333333333333e-06,
"loss": 0.0024,
"step": 7300
},
{
"epoch": 42.84,
"learning_rate": 6.500512820512821e-06,
"loss": 0.0025,
"step": 7325
},
{
"epoch": 42.98,
"learning_rate": 6.487692307692308e-06,
"loss": 0.0023,
"step": 7350
},
{
"epoch": 43.13,
"learning_rate": 6.4748717948717955e-06,
"loss": 0.0028,
"step": 7375
},
{
"epoch": 43.27,
"learning_rate": 6.462051282051282e-06,
"loss": 0.0026,
"step": 7400
},
{
"epoch": 43.42,
"learning_rate": 6.44923076923077e-06,
"loss": 0.0027,
"step": 7425
},
{
"epoch": 43.57,
"learning_rate": 6.436410256410257e-06,
"loss": 0.0027,
"step": 7450
},
{
"epoch": 43.71,
"learning_rate": 6.4235897435897445e-06,
"loss": 0.0029,
"step": 7475
},
{
"epoch": 43.86,
"learning_rate": 6.410769230769231e-06,
"loss": 0.0022,
"step": 7500
},
{
"epoch": 44.01,
"learning_rate": 6.397948717948719e-06,
"loss": 0.0025,
"step": 7525
},
{
"epoch": 44.15,
"learning_rate": 6.385128205128206e-06,
"loss": 0.0021,
"step": 7550
},
{
"epoch": 44.3,
"learning_rate": 6.3723076923076935e-06,
"loss": 0.0023,
"step": 7575
},
{
"epoch": 44.44,
"learning_rate": 6.35948717948718e-06,
"loss": 0.0022,
"step": 7600
},
{
"epoch": 44.59,
"learning_rate": 6.346666666666668e-06,
"loss": 0.0016,
"step": 7625
},
{
"epoch": 44.74,
"learning_rate": 6.333846153846155e-06,
"loss": 0.0018,
"step": 7650
},
{
"epoch": 44.88,
"learning_rate": 6.3210256410256425e-06,
"loss": 0.002,
"step": 7675
},
{
"epoch": 45.03,
"learning_rate": 6.3082051282051285e-06,
"loss": 0.0017,
"step": 7700
},
{
"epoch": 45.18,
"learning_rate": 6.295384615384615e-06,
"loss": 0.0024,
"step": 7725
},
{
"epoch": 45.32,
"learning_rate": 6.282564102564103e-06,
"loss": 0.0028,
"step": 7750
},
{
"epoch": 45.47,
"learning_rate": 6.26974358974359e-06,
"loss": 0.0026,
"step": 7775
},
{
"epoch": 45.61,
"learning_rate": 6.2569230769230775e-06,
"loss": 0.0024,
"step": 7800
},
{
"epoch": 45.76,
"learning_rate": 6.244102564102564e-06,
"loss": 0.0021,
"step": 7825
},
{
"epoch": 45.91,
"learning_rate": 6.231282051282051e-06,
"loss": 0.0021,
"step": 7850
},
{
"epoch": 46.05,
"learning_rate": 6.218461538461539e-06,
"loss": 0.0022,
"step": 7875
},
{
"epoch": 46.2,
"learning_rate": 6.205641025641026e-06,
"loss": 0.0016,
"step": 7900
},
{
"epoch": 46.35,
"learning_rate": 6.192820512820513e-06,
"loss": 0.0021,
"step": 7925
},
{
"epoch": 46.49,
"learning_rate": 6.18e-06,
"loss": 0.0028,
"step": 7950
},
{
"epoch": 46.64,
"learning_rate": 6.167179487179488e-06,
"loss": 0.003,
"step": 7975
},
{
"epoch": 46.78,
"learning_rate": 6.154358974358975e-06,
"loss": 0.0023,
"step": 8000
},
{
"epoch": 46.78,
"eval_loss": 0.3455485999584198,
"eval_runtime": 1416.7263,
"eval_samples_per_second": 4.652,
"eval_steps_per_second": 0.291,
"eval_wer": 13.848757782747988,
"step": 8000
},
{
"epoch": 46.93,
"learning_rate": 6.141538461538462e-06,
"loss": 0.0018,
"step": 8025
},
{
"epoch": 47.08,
"learning_rate": 6.128717948717949e-06,
"loss": 0.0015,
"step": 8050
},
{
"epoch": 47.22,
"learning_rate": 6.115897435897437e-06,
"loss": 0.0022,
"step": 8075
},
{
"epoch": 47.37,
"learning_rate": 6.103076923076924e-06,
"loss": 0.0024,
"step": 8100
},
{
"epoch": 47.51,
"learning_rate": 6.090256410256411e-06,
"loss": 0.0017,
"step": 8125
},
{
"epoch": 47.66,
"learning_rate": 6.077435897435898e-06,
"loss": 0.0016,
"step": 8150
},
{
"epoch": 47.81,
"learning_rate": 6.064615384615386e-06,
"loss": 0.0016,
"step": 8175
},
{
"epoch": 47.95,
"learning_rate": 6.051794871794873e-06,
"loss": 0.0021,
"step": 8200
},
{
"epoch": 48.1,
"learning_rate": 6.03897435897436e-06,
"loss": 0.0017,
"step": 8225
},
{
"epoch": 48.25,
"learning_rate": 6.026153846153847e-06,
"loss": 0.0018,
"step": 8250
},
{
"epoch": 48.39,
"learning_rate": 6.013333333333335e-06,
"loss": 0.0017,
"step": 8275
},
{
"epoch": 48.54,
"learning_rate": 6.000512820512821e-06,
"loss": 0.0024,
"step": 8300
},
{
"epoch": 48.68,
"learning_rate": 5.9876923076923076e-06,
"loss": 0.0021,
"step": 8325
},
{
"epoch": 48.83,
"learning_rate": 5.974871794871795e-06,
"loss": 0.0023,
"step": 8350
},
{
"epoch": 48.98,
"learning_rate": 5.962051282051282e-06,
"loss": 0.0027,
"step": 8375
},
{
"epoch": 49.12,
"learning_rate": 5.94923076923077e-06,
"loss": 0.0035,
"step": 8400
},
{
"epoch": 49.27,
"learning_rate": 5.9364102564102566e-06,
"loss": 0.0024,
"step": 8425
},
{
"epoch": 49.42,
"learning_rate": 5.923589743589744e-06,
"loss": 0.0026,
"step": 8450
},
{
"epoch": 49.56,
"learning_rate": 5.910769230769231e-06,
"loss": 0.0018,
"step": 8475
},
{
"epoch": 49.71,
"learning_rate": 5.897948717948718e-06,
"loss": 0.0018,
"step": 8500
},
{
"epoch": 49.85,
"learning_rate": 5.8851282051282056e-06,
"loss": 0.0018,
"step": 8525
},
{
"epoch": 50.0,
"learning_rate": 5.872307692307692e-06,
"loss": 0.0031,
"step": 8550
},
{
"epoch": 50.15,
"learning_rate": 5.85948717948718e-06,
"loss": 0.0018,
"step": 8575
},
{
"epoch": 50.29,
"learning_rate": 5.846666666666667e-06,
"loss": 0.0021,
"step": 8600
},
{
"epoch": 50.44,
"learning_rate": 5.8338461538461545e-06,
"loss": 0.0017,
"step": 8625
},
{
"epoch": 50.58,
"learning_rate": 5.821025641025641e-06,
"loss": 0.0013,
"step": 8650
},
{
"epoch": 50.73,
"learning_rate": 5.808205128205129e-06,
"loss": 0.0018,
"step": 8675
},
{
"epoch": 50.88,
"learning_rate": 5.795384615384616e-06,
"loss": 0.0014,
"step": 8700
},
{
"epoch": 51.02,
"learning_rate": 5.7825641025641035e-06,
"loss": 0.0021,
"step": 8725
},
{
"epoch": 51.17,
"learning_rate": 5.76974358974359e-06,
"loss": 0.0019,
"step": 8750
},
{
"epoch": 51.32,
"learning_rate": 5.756923076923078e-06,
"loss": 0.0018,
"step": 8775
},
{
"epoch": 51.46,
"learning_rate": 5.744102564102565e-06,
"loss": 0.0024,
"step": 8800
},
{
"epoch": 51.61,
"learning_rate": 5.7312820512820525e-06,
"loss": 0.0027,
"step": 8825
},
{
"epoch": 51.75,
"learning_rate": 5.718461538461539e-06,
"loss": 0.0028,
"step": 8850
},
{
"epoch": 51.9,
"learning_rate": 5.705641025641027e-06,
"loss": 0.0024,
"step": 8875
},
{
"epoch": 52.05,
"learning_rate": 5.692820512820513e-06,
"loss": 0.0027,
"step": 8900
},
{
"epoch": 52.19,
"learning_rate": 5.68e-06,
"loss": 0.0016,
"step": 8925
},
{
"epoch": 52.34,
"learning_rate": 5.6671794871794875e-06,
"loss": 0.0015,
"step": 8950
},
{
"epoch": 52.49,
"learning_rate": 5.654358974358974e-06,
"loss": 0.0027,
"step": 8975
},
{
"epoch": 52.63,
"learning_rate": 5.641538461538462e-06,
"loss": 0.002,
"step": 9000
},
{
"epoch": 52.63,
"eval_loss": 0.3499505817890167,
"eval_runtime": 1411.8964,
"eval_samples_per_second": 4.668,
"eval_steps_per_second": 0.292,
"eval_wer": 13.512260976445223,
"step": 9000
},
{
"epoch": 52.78,
"learning_rate": 5.628717948717949e-06,
"loss": 0.0018,
"step": 9025
},
{
"epoch": 52.92,
"learning_rate": 5.6158974358974365e-06,
"loss": 0.0014,
"step": 9050
},
{
"epoch": 53.07,
"learning_rate": 5.603076923076923e-06,
"loss": 0.0021,
"step": 9075
},
{
"epoch": 53.22,
"learning_rate": 5.59025641025641e-06,
"loss": 0.0018,
"step": 9100
},
{
"epoch": 53.36,
"learning_rate": 5.577435897435898e-06,
"loss": 0.0015,
"step": 9125
},
{
"epoch": 53.51,
"learning_rate": 5.564615384615385e-06,
"loss": 0.0014,
"step": 9150
},
{
"epoch": 53.65,
"learning_rate": 5.551794871794872e-06,
"loss": 0.0021,
"step": 9175
},
{
"epoch": 53.8,
"learning_rate": 5.538974358974359e-06,
"loss": 0.0014,
"step": 9200
},
{
"epoch": 53.95,
"learning_rate": 5.526153846153847e-06,
"loss": 0.0014,
"step": 9225
},
{
"epoch": 54.09,
"learning_rate": 5.513333333333334e-06,
"loss": 0.0015,
"step": 9250
},
{
"epoch": 54.24,
"learning_rate": 5.500512820512821e-06,
"loss": 0.0022,
"step": 9275
},
{
"epoch": 54.39,
"learning_rate": 5.487692307692308e-06,
"loss": 0.0014,
"step": 9300
},
{
"epoch": 54.53,
"learning_rate": 5.474871794871796e-06,
"loss": 0.0012,
"step": 9325
},
{
"epoch": 54.68,
"learning_rate": 5.462051282051283e-06,
"loss": 0.0014,
"step": 9350
},
{
"epoch": 54.82,
"learning_rate": 5.44923076923077e-06,
"loss": 0.0019,
"step": 9375
},
{
"epoch": 54.97,
"learning_rate": 5.436410256410257e-06,
"loss": 0.002,
"step": 9400
},
{
"epoch": 55.12,
"learning_rate": 5.423589743589745e-06,
"loss": 0.002,
"step": 9425
},
{
"epoch": 55.26,
"learning_rate": 5.410769230769232e-06,
"loss": 0.0013,
"step": 9450
},
{
"epoch": 55.41,
"learning_rate": 5.397948717948719e-06,
"loss": 0.0014,
"step": 9475
},
{
"epoch": 55.56,
"learning_rate": 5.385128205128206e-06,
"loss": 0.0016,
"step": 9500
},
{
"epoch": 55.7,
"learning_rate": 5.372307692307692e-06,
"loss": 0.0022,
"step": 9525
},
{
"epoch": 55.85,
"learning_rate": 5.35948717948718e-06,
"loss": 0.0021,
"step": 9550
},
{
"epoch": 55.99,
"learning_rate": 5.346666666666667e-06,
"loss": 0.0021,
"step": 9575
},
{
"epoch": 56.14,
"learning_rate": 5.333846153846154e-06,
"loss": 0.0014,
"step": 9600
},
{
"epoch": 56.29,
"learning_rate": 5.321025641025641e-06,
"loss": 0.001,
"step": 9625
},
{
"epoch": 56.43,
"learning_rate": 5.308205128205129e-06,
"loss": 0.0015,
"step": 9650
},
{
"epoch": 56.58,
"learning_rate": 5.2953846153846156e-06,
"loss": 0.0013,
"step": 9675
},
{
"epoch": 56.73,
"learning_rate": 5.282564102564102e-06,
"loss": 0.0017,
"step": 9700
},
{
"epoch": 56.87,
"learning_rate": 5.26974358974359e-06,
"loss": 0.0013,
"step": 9725
},
{
"epoch": 57.02,
"learning_rate": 5.256923076923077e-06,
"loss": 0.001,
"step": 9750
},
{
"epoch": 57.16,
"learning_rate": 5.2441025641025646e-06,
"loss": 0.0014,
"step": 9775
},
{
"epoch": 57.31,
"learning_rate": 5.231282051282051e-06,
"loss": 0.0011,
"step": 9800
},
{
"epoch": 57.46,
"learning_rate": 5.218461538461539e-06,
"loss": 0.001,
"step": 9825
},
{
"epoch": 57.6,
"learning_rate": 5.205641025641026e-06,
"loss": 0.0009,
"step": 9850
},
{
"epoch": 57.75,
"learning_rate": 5.1928205128205136e-06,
"loss": 0.001,
"step": 9875
},
{
"epoch": 57.89,
"learning_rate": 5.18e-06,
"loss": 0.001,
"step": 9900
},
{
"epoch": 58.04,
"learning_rate": 5.167179487179488e-06,
"loss": 0.0009,
"step": 9925
},
{
"epoch": 58.19,
"learning_rate": 5.154358974358975e-06,
"loss": 0.0008,
"step": 9950
},
{
"epoch": 58.33,
"learning_rate": 5.1415384615384625e-06,
"loss": 0.0009,
"step": 9975
},
{
"epoch": 58.48,
"learning_rate": 5.128717948717949e-06,
"loss": 0.0009,
"step": 10000
},
{
"epoch": 58.48,
"eval_loss": 0.35903850197792053,
"eval_runtime": 1412.1699,
"eval_samples_per_second": 4.667,
"eval_steps_per_second": 0.292,
"eval_wer": 13.29666122629914,
"step": 10000
},
{
"epoch": 58.63,
"learning_rate": 5.115897435897437e-06,
"loss": 0.0011,
"step": 10025
},
{
"epoch": 58.77,
"learning_rate": 5.103076923076924e-06,
"loss": 0.0007,
"step": 10050
},
{
"epoch": 58.92,
"learning_rate": 5.0902564102564115e-06,
"loss": 0.0009,
"step": 10075
},
{
"epoch": 59.06,
"learning_rate": 5.077948717948719e-06,
"loss": 0.0013,
"step": 10100
},
{
"epoch": 59.21,
"learning_rate": 5.065128205128206e-06,
"loss": 0.0013,
"step": 10125
},
{
"epoch": 59.36,
"learning_rate": 5.052307692307693e-06,
"loss": 0.0015,
"step": 10150
},
{
"epoch": 59.5,
"learning_rate": 5.03948717948718e-06,
"loss": 0.0012,
"step": 10175
},
{
"epoch": 59.65,
"learning_rate": 5.026666666666667e-06,
"loss": 0.0015,
"step": 10200
},
{
"epoch": 59.8,
"learning_rate": 5.013846153846155e-06,
"loss": 0.0014,
"step": 10225
},
{
"epoch": 59.94,
"learning_rate": 5.001025641025642e-06,
"loss": 0.0019,
"step": 10250
},
{
"epoch": 60.09,
"learning_rate": 4.988205128205128e-06,
"loss": 0.0015,
"step": 10275
},
{
"epoch": 60.23,
"learning_rate": 4.975384615384616e-06,
"loss": 0.0014,
"step": 10300
},
{
"epoch": 60.38,
"learning_rate": 4.962564102564103e-06,
"loss": 0.0014,
"step": 10325
},
{
"epoch": 60.53,
"learning_rate": 4.949743589743591e-06,
"loss": 0.0012,
"step": 10350
},
{
"epoch": 60.67,
"learning_rate": 4.936923076923077e-06,
"loss": 0.0012,
"step": 10375
},
{
"epoch": 60.82,
"learning_rate": 4.924102564102565e-06,
"loss": 0.001,
"step": 10400
},
{
"epoch": 60.96,
"learning_rate": 4.911282051282052e-06,
"loss": 0.0012,
"step": 10425
},
{
"epoch": 61.11,
"learning_rate": 4.898461538461539e-06,
"loss": 0.0009,
"step": 10450
},
{
"epoch": 61.26,
"learning_rate": 4.8856410256410256e-06,
"loss": 0.0008,
"step": 10475
},
{
"epoch": 61.4,
"learning_rate": 4.872820512820513e-06,
"loss": 0.0008,
"step": 10500
},
{
"epoch": 61.55,
"learning_rate": 4.86e-06,
"loss": 0.0009,
"step": 10525
},
{
"epoch": 61.7,
"learning_rate": 4.847179487179488e-06,
"loss": 0.0017,
"step": 10550
},
{
"epoch": 61.84,
"learning_rate": 4.8343589743589746e-06,
"loss": 0.0013,
"step": 10575
},
{
"epoch": 61.99,
"learning_rate": 4.821538461538462e-06,
"loss": 0.0017,
"step": 10600
},
{
"epoch": 62.13,
"learning_rate": 4.808717948717949e-06,
"loss": 0.0021,
"step": 10625
},
{
"epoch": 62.28,
"learning_rate": 4.795897435897437e-06,
"loss": 0.002,
"step": 10650
},
{
"epoch": 62.43,
"learning_rate": 4.7830769230769235e-06,
"loss": 0.0016,
"step": 10675
},
{
"epoch": 62.57,
"learning_rate": 4.770256410256411e-06,
"loss": 0.0011,
"step": 10700
},
{
"epoch": 62.72,
"learning_rate": 4.757435897435898e-06,
"loss": 0.0012,
"step": 10725
},
{
"epoch": 62.87,
"learning_rate": 4.744615384615385e-06,
"loss": 0.0014,
"step": 10750
},
{
"epoch": 63.01,
"learning_rate": 4.731794871794872e-06,
"loss": 0.0017,
"step": 10775
},
{
"epoch": 63.16,
"learning_rate": 4.718974358974359e-06,
"loss": 0.0014,
"step": 10800
},
{
"epoch": 63.3,
"learning_rate": 4.706153846153846e-06,
"loss": 0.0013,
"step": 10825
},
{
"epoch": 63.45,
"learning_rate": 4.693333333333334e-06,
"loss": 0.0008,
"step": 10850
},
{
"epoch": 63.6,
"learning_rate": 4.680512820512821e-06,
"loss": 0.0011,
"step": 10875
},
{
"epoch": 63.74,
"learning_rate": 4.667692307692308e-06,
"loss": 0.0016,
"step": 10900
},
{
"epoch": 63.89,
"learning_rate": 4.654871794871795e-06,
"loss": 0.0015,
"step": 10925
},
{
"epoch": 64.04,
"learning_rate": 4.642051282051283e-06,
"loss": 0.0018,
"step": 10950
},
{
"epoch": 64.18,
"learning_rate": 4.62923076923077e-06,
"loss": 0.0013,
"step": 10975
},
{
"epoch": 64.33,
"learning_rate": 4.616410256410257e-06,
"loss": 0.0016,
"step": 11000
},
{
"epoch": 64.33,
"eval_loss": 0.3674701452255249,
"eval_runtime": 1413.2133,
"eval_samples_per_second": 4.664,
"eval_steps_per_second": 0.292,
"eval_wer": 13.46793205585444,
"step": 11000
},
{
"epoch": 64.47,
"learning_rate": 4.603589743589744e-06,
"loss": 0.0013,
"step": 11025
},
{
"epoch": 64.62,
"learning_rate": 4.590769230769232e-06,
"loss": 0.0013,
"step": 11050
},
{
"epoch": 64.77,
"learning_rate": 4.577948717948718e-06,
"loss": 0.0012,
"step": 11075
},
{
"epoch": 64.91,
"learning_rate": 4.5651282051282055e-06,
"loss": 0.0013,
"step": 11100
},
{
"epoch": 65.06,
"learning_rate": 4.552307692307692e-06,
"loss": 0.0012,
"step": 11125
},
{
"epoch": 65.2,
"learning_rate": 4.53948717948718e-06,
"loss": 0.0008,
"step": 11150
},
{
"epoch": 65.35,
"learning_rate": 4.526666666666667e-06,
"loss": 0.0008,
"step": 11175
},
{
"epoch": 65.5,
"learning_rate": 4.5138461538461545e-06,
"loss": 0.0008,
"step": 11200
},
{
"epoch": 65.64,
"learning_rate": 4.501025641025641e-06,
"loss": 0.0011,
"step": 11225
},
{
"epoch": 65.79,
"learning_rate": 4.488205128205129e-06,
"loss": 0.0008,
"step": 11250
},
{
"epoch": 65.94,
"learning_rate": 4.475384615384616e-06,
"loss": 0.0008,
"step": 11275
},
{
"epoch": 66.08,
"learning_rate": 4.4625641025641035e-06,
"loss": 0.0007,
"step": 11300
},
{
"epoch": 66.23,
"learning_rate": 4.44974358974359e-06,
"loss": 0.0006,
"step": 11325
},
{
"epoch": 66.37,
"learning_rate": 4.436923076923078e-06,
"loss": 0.001,
"step": 11350
},
{
"epoch": 66.52,
"learning_rate": 4.424102564102564e-06,
"loss": 0.001,
"step": 11375
},
{
"epoch": 66.67,
"learning_rate": 4.411282051282052e-06,
"loss": 0.0017,
"step": 11400
},
{
"epoch": 66.81,
"learning_rate": 4.3984615384615384e-06,
"loss": 0.0009,
"step": 11425
},
{
"epoch": 66.96,
"learning_rate": 4.385641025641026e-06,
"loss": 0.0007,
"step": 11450
},
{
"epoch": 67.11,
"learning_rate": 4.372820512820513e-06,
"loss": 0.0006,
"step": 11475
},
{
"epoch": 67.25,
"learning_rate": 4.360000000000001e-06,
"loss": 0.0008,
"step": 11500
},
{
"epoch": 67.4,
"learning_rate": 4.3471794871794874e-06,
"loss": 0.0006,
"step": 11525
},
{
"epoch": 67.54,
"learning_rate": 4.334358974358975e-06,
"loss": 0.0005,
"step": 11550
},
{
"epoch": 67.69,
"learning_rate": 4.321538461538462e-06,
"loss": 0.0007,
"step": 11575
},
{
"epoch": 67.84,
"learning_rate": 4.30871794871795e-06,
"loss": 0.0005,
"step": 11600
},
{
"epoch": 67.98,
"learning_rate": 4.2958974358974364e-06,
"loss": 0.0007,
"step": 11625
},
{
"epoch": 68.13,
"learning_rate": 4.283076923076924e-06,
"loss": 0.0006,
"step": 11650
},
{
"epoch": 68.27,
"learning_rate": 4.27025641025641e-06,
"loss": 0.0005,
"step": 11675
},
{
"epoch": 68.42,
"learning_rate": 4.257435897435898e-06,
"loss": 0.0009,
"step": 11700
},
{
"epoch": 68.57,
"learning_rate": 4.2446153846153846e-06,
"loss": 0.0006,
"step": 11725
},
{
"epoch": 68.71,
"learning_rate": 4.231794871794872e-06,
"loss": 0.0005,
"step": 11750
},
{
"epoch": 68.86,
"learning_rate": 4.218974358974359e-06,
"loss": 0.0006,
"step": 11775
},
{
"epoch": 69.01,
"learning_rate": 4.206153846153847e-06,
"loss": 0.0008,
"step": 11800
},
{
"epoch": 69.15,
"learning_rate": 4.1933333333333336e-06,
"loss": 0.0012,
"step": 11825
},
{
"epoch": 69.3,
"learning_rate": 4.180512820512821e-06,
"loss": 0.0011,
"step": 11850
},
{
"epoch": 69.44,
"learning_rate": 4.167692307692308e-06,
"loss": 0.0013,
"step": 11875
},
{
"epoch": 69.59,
"learning_rate": 4.154871794871796e-06,
"loss": 0.0009,
"step": 11900
},
{
"epoch": 69.74,
"learning_rate": 4.1420512820512826e-06,
"loss": 0.0008,
"step": 11925
},
{
"epoch": 69.88,
"learning_rate": 4.12923076923077e-06,
"loss": 0.0007,
"step": 11950
},
{
"epoch": 70.03,
"learning_rate": 4.116410256410256e-06,
"loss": 0.0007,
"step": 11975
},
{
"epoch": 70.18,
"learning_rate": 4.103589743589744e-06,
"loss": 0.0007,
"step": 12000
},
{
"epoch": 70.18,
"eval_loss": 0.37850871682167053,
"eval_runtime": 1414.6429,
"eval_samples_per_second": 4.659,
"eval_steps_per_second": 0.291,
"eval_wer": 13.268451913195914,
"step": 12000
},
{
"epoch": 70.32,
"learning_rate": 4.090769230769231e-06,
"loss": 0.0007,
"step": 12025
},
{
"epoch": 70.47,
"learning_rate": 4.077948717948718e-06,
"loss": 0.0006,
"step": 12050
},
{
"epoch": 70.61,
"learning_rate": 4.065128205128205e-06,
"loss": 0.001,
"step": 12075
},
{
"epoch": 70.76,
"learning_rate": 4.052307692307693e-06,
"loss": 0.0006,
"step": 12100
},
{
"epoch": 70.91,
"learning_rate": 4.03948717948718e-06,
"loss": 0.0006,
"step": 12125
},
{
"epoch": 71.05,
"learning_rate": 4.026666666666667e-06,
"loss": 0.0008,
"step": 12150
},
{
"epoch": 71.2,
"learning_rate": 4.013846153846154e-06,
"loss": 0.0009,
"step": 12175
},
{
"epoch": 71.35,
"learning_rate": 4.001025641025642e-06,
"loss": 0.0007,
"step": 12200
},
{
"epoch": 71.49,
"learning_rate": 3.988205128205129e-06,
"loss": 0.0007,
"step": 12225
},
{
"epoch": 71.64,
"learning_rate": 3.975384615384616e-06,
"loss": 0.0009,
"step": 12250
},
{
"epoch": 71.78,
"learning_rate": 3.962564102564102e-06,
"loss": 0.0013,
"step": 12275
},
{
"epoch": 71.93,
"learning_rate": 3.94974358974359e-06,
"loss": 0.0008,
"step": 12300
},
{
"epoch": 72.08,
"learning_rate": 3.936923076923077e-06,
"loss": 0.0006,
"step": 12325
},
{
"epoch": 72.22,
"learning_rate": 3.9241025641025645e-06,
"loss": 0.0007,
"step": 12350
},
{
"epoch": 72.37,
"learning_rate": 3.911282051282051e-06,
"loss": 0.0006,
"step": 12375
},
{
"epoch": 72.51,
"learning_rate": 3.898461538461539e-06,
"loss": 0.0006,
"step": 12400
},
{
"epoch": 72.66,
"learning_rate": 3.885641025641026e-06,
"loss": 0.0005,
"step": 12425
},
{
"epoch": 72.81,
"learning_rate": 3.8728205128205135e-06,
"loss": 0.0006,
"step": 12450
},
{
"epoch": 72.95,
"learning_rate": 3.86e-06,
"loss": 0.0006,
"step": 12475
},
{
"epoch": 73.1,
"learning_rate": 3.847179487179488e-06,
"loss": 0.0006,
"step": 12500
},
{
"epoch": 73.25,
"learning_rate": 3.834358974358975e-06,
"loss": 0.0008,
"step": 12525
},
{
"epoch": 73.39,
"learning_rate": 3.8215384615384625e-06,
"loss": 0.0007,
"step": 12550
},
{
"epoch": 73.54,
"learning_rate": 3.808717948717949e-06,
"loss": 0.0005,
"step": 12575
},
{
"epoch": 73.68,
"learning_rate": 3.795897435897436e-06,
"loss": 0.0006,
"step": 12600
},
{
"epoch": 73.83,
"learning_rate": 3.7830769230769234e-06,
"loss": 0.0008,
"step": 12625
},
{
"epoch": 73.98,
"learning_rate": 3.7702564102564106e-06,
"loss": 0.0014,
"step": 12650
},
{
"epoch": 74.12,
"learning_rate": 3.7574358974358974e-06,
"loss": 0.0011,
"step": 12675
},
{
"epoch": 74.27,
"learning_rate": 3.7446153846153847e-06,
"loss": 0.001,
"step": 12700
},
{
"epoch": 74.42,
"learning_rate": 3.731794871794872e-06,
"loss": 0.0008,
"step": 12725
},
{
"epoch": 74.56,
"learning_rate": 3.718974358974359e-06,
"loss": 0.0007,
"step": 12750
},
{
"epoch": 74.71,
"learning_rate": 3.7061538461538464e-06,
"loss": 0.0013,
"step": 12775
},
{
"epoch": 74.85,
"learning_rate": 3.6933333333333337e-06,
"loss": 0.0008,
"step": 12800
},
{
"epoch": 75.0,
"learning_rate": 3.680512820512821e-06,
"loss": 0.001,
"step": 12825
},
{
"epoch": 75.15,
"learning_rate": 3.667692307692308e-06,
"loss": 0.0006,
"step": 12850
},
{
"epoch": 75.29,
"learning_rate": 3.6548717948717954e-06,
"loss": 0.0006,
"step": 12875
},
{
"epoch": 75.44,
"learning_rate": 3.6420512820512823e-06,
"loss": 0.0007,
"step": 12900
},
{
"epoch": 75.58,
"learning_rate": 3.6292307692307695e-06,
"loss": 0.0012,
"step": 12925
},
{
"epoch": 75.73,
"learning_rate": 3.6164102564102568e-06,
"loss": 0.0009,
"step": 12950
},
{
"epoch": 75.88,
"learning_rate": 3.6035897435897436e-06,
"loss": 0.0007,
"step": 12975
},
{
"epoch": 76.02,
"learning_rate": 3.590769230769231e-06,
"loss": 0.0008,
"step": 13000
},
{
"epoch": 76.02,
"eval_loss": 0.38218748569488525,
"eval_runtime": 1413.7781,
"eval_samples_per_second": 4.662,
"eval_steps_per_second": 0.291,
"eval_wer": 13.36516955812126,
"step": 13000
},
{
"epoch": 76.17,
"learning_rate": 3.577948717948718e-06,
"loss": 0.0011,
"step": 13025
},
{
"epoch": 76.32,
"learning_rate": 3.5651282051282053e-06,
"loss": 0.0005,
"step": 13050
},
{
"epoch": 76.46,
"learning_rate": 3.5523076923076926e-06,
"loss": 0.0005,
"step": 13075
},
{
"epoch": 76.61,
"learning_rate": 3.53948717948718e-06,
"loss": 0.0005,
"step": 13100
},
{
"epoch": 76.75,
"learning_rate": 3.526666666666667e-06,
"loss": 0.0006,
"step": 13125
},
{
"epoch": 76.9,
"learning_rate": 3.5138461538461543e-06,
"loss": 0.0007,
"step": 13150
},
{
"epoch": 77.05,
"learning_rate": 3.5015384615384617e-06,
"loss": 0.0007,
"step": 13175
},
{
"epoch": 77.19,
"learning_rate": 3.488717948717949e-06,
"loss": 0.0013,
"step": 13200
},
{
"epoch": 77.34,
"learning_rate": 3.4758974358974362e-06,
"loss": 0.0007,
"step": 13225
},
{
"epoch": 77.49,
"learning_rate": 3.4630769230769235e-06,
"loss": 0.0009,
"step": 13250
},
{
"epoch": 77.63,
"learning_rate": 3.4502564102564107e-06,
"loss": 0.0005,
"step": 13275
},
{
"epoch": 77.78,
"learning_rate": 3.437435897435898e-06,
"loss": 0.0005,
"step": 13300
},
{
"epoch": 77.92,
"learning_rate": 3.4246153846153852e-06,
"loss": 0.0007,
"step": 13325
},
{
"epoch": 78.07,
"learning_rate": 3.4117948717948725e-06,
"loss": 0.0005,
"step": 13350
},
{
"epoch": 78.22,
"learning_rate": 3.398974358974359e-06,
"loss": 0.0004,
"step": 13375
},
{
"epoch": 78.36,
"learning_rate": 3.386153846153846e-06,
"loss": 0.0005,
"step": 13400
},
{
"epoch": 78.51,
"learning_rate": 3.3733333333333334e-06,
"loss": 0.0007,
"step": 13425
},
{
"epoch": 78.65,
"learning_rate": 3.3605128205128206e-06,
"loss": 0.0008,
"step": 13450
},
{
"epoch": 78.8,
"learning_rate": 3.347692307692308e-06,
"loss": 0.0004,
"step": 13475
},
{
"epoch": 78.95,
"learning_rate": 3.334871794871795e-06,
"loss": 0.0005,
"step": 13500
},
{
"epoch": 79.09,
"learning_rate": 3.3220512820512824e-06,
"loss": 0.0007,
"step": 13525
},
{
"epoch": 79.24,
"learning_rate": 3.3092307692307696e-06,
"loss": 0.0004,
"step": 13550
},
{
"epoch": 79.39,
"learning_rate": 3.296410256410257e-06,
"loss": 0.0007,
"step": 13575
},
{
"epoch": 79.53,
"learning_rate": 3.283589743589744e-06,
"loss": 0.0005,
"step": 13600
},
{
"epoch": 79.68,
"learning_rate": 3.2707692307692313e-06,
"loss": 0.0004,
"step": 13625
},
{
"epoch": 79.82,
"learning_rate": 3.2579487179487186e-06,
"loss": 0.0005,
"step": 13650
},
{
"epoch": 79.97,
"learning_rate": 3.245128205128205e-06,
"loss": 0.0006,
"step": 13675
},
{
"epoch": 80.12,
"learning_rate": 3.2323076923076922e-06,
"loss": 0.0005,
"step": 13700
},
{
"epoch": 80.26,
"learning_rate": 3.2194871794871795e-06,
"loss": 0.0004,
"step": 13725
},
{
"epoch": 80.41,
"learning_rate": 3.2066666666666667e-06,
"loss": 0.0006,
"step": 13750
},
{
"epoch": 80.56,
"learning_rate": 3.193846153846154e-06,
"loss": 0.0006,
"step": 13775
},
{
"epoch": 80.7,
"learning_rate": 3.1810256410256412e-06,
"loss": 0.0004,
"step": 13800
},
{
"epoch": 80.85,
"learning_rate": 3.1682051282051285e-06,
"loss": 0.0004,
"step": 13825
},
{
"epoch": 80.99,
"learning_rate": 3.1553846153846157e-06,
"loss": 0.0008,
"step": 13850
},
{
"epoch": 81.14,
"learning_rate": 3.142564102564103e-06,
"loss": 0.0005,
"step": 13875
},
{
"epoch": 81.29,
"learning_rate": 3.1297435897435902e-06,
"loss": 0.0005,
"step": 13900
},
{
"epoch": 81.43,
"learning_rate": 3.1169230769230775e-06,
"loss": 0.0006,
"step": 13925
},
{
"epoch": 81.58,
"learning_rate": 3.1041025641025647e-06,
"loss": 0.0005,
"step": 13950
},
{
"epoch": 81.73,
"learning_rate": 3.091282051282051e-06,
"loss": 0.0006,
"step": 13975
},
{
"epoch": 81.87,
"learning_rate": 3.0784615384615384e-06,
"loss": 0.0004,
"step": 14000
},
{
"epoch": 81.87,
"eval_loss": 0.39293619990348816,
"eval_runtime": 1411.9422,
"eval_samples_per_second": 4.668,
"eval_steps_per_second": 0.292,
"eval_wer": 13.31479578472264,
"step": 14000
},
{
"epoch": 82.02,
"learning_rate": 3.0656410256410256e-06,
"loss": 0.0004,
"step": 14025
},
{
"epoch": 82.16,
"learning_rate": 3.052820512820513e-06,
"loss": 0.0005,
"step": 14050
},
{
"epoch": 82.31,
"learning_rate": 3.04e-06,
"loss": 0.0005,
"step": 14075
},
{
"epoch": 82.46,
"learning_rate": 3.0271794871794874e-06,
"loss": 0.0007,
"step": 14100
},
{
"epoch": 82.6,
"learning_rate": 3.0143589743589746e-06,
"loss": 0.0005,
"step": 14125
},
{
"epoch": 82.75,
"learning_rate": 3.001538461538462e-06,
"loss": 0.0005,
"step": 14150
},
{
"epoch": 82.89,
"learning_rate": 2.988717948717949e-06,
"loss": 0.0004,
"step": 14175
},
{
"epoch": 83.04,
"learning_rate": 2.9758974358974364e-06,
"loss": 0.0004,
"step": 14200
},
{
"epoch": 83.19,
"learning_rate": 2.9630769230769236e-06,
"loss": 0.0004,
"step": 14225
},
{
"epoch": 83.33,
"learning_rate": 2.950256410256411e-06,
"loss": 0.0005,
"step": 14250
},
{
"epoch": 83.48,
"learning_rate": 2.9374358974358973e-06,
"loss": 0.0005,
"step": 14275
},
{
"epoch": 83.63,
"learning_rate": 2.9246153846153845e-06,
"loss": 0.0007,
"step": 14300
},
{
"epoch": 83.77,
"learning_rate": 2.9117948717948717e-06,
"loss": 0.0006,
"step": 14325
},
{
"epoch": 83.92,
"learning_rate": 2.898974358974359e-06,
"loss": 0.0008,
"step": 14350
},
{
"epoch": 84.06,
"learning_rate": 2.8861538461538462e-06,
"loss": 0.0006,
"step": 14375
},
{
"epoch": 84.21,
"learning_rate": 2.8733333333333335e-06,
"loss": 0.0006,
"step": 14400
},
{
"epoch": 84.36,
"learning_rate": 2.8605128205128207e-06,
"loss": 0.0008,
"step": 14425
},
{
"epoch": 84.5,
"learning_rate": 2.847692307692308e-06,
"loss": 0.0004,
"step": 14450
},
{
"epoch": 84.65,
"learning_rate": 2.8348717948717952e-06,
"loss": 0.0004,
"step": 14475
},
{
"epoch": 84.8,
"learning_rate": 2.8220512820512825e-06,
"loss": 0.0003,
"step": 14500
},
{
"epoch": 84.94,
"learning_rate": 2.8092307692307697e-06,
"loss": 0.0003,
"step": 14525
},
{
"epoch": 85.09,
"learning_rate": 2.796410256410257e-06,
"loss": 0.0004,
"step": 14550
},
{
"epoch": 85.23,
"learning_rate": 2.7835897435897434e-06,
"loss": 0.0004,
"step": 14575
},
{
"epoch": 85.38,
"learning_rate": 2.7707692307692306e-06,
"loss": 0.0003,
"step": 14600
},
{
"epoch": 85.53,
"learning_rate": 2.757948717948718e-06,
"loss": 0.0003,
"step": 14625
},
{
"epoch": 85.67,
"learning_rate": 2.745128205128205e-06,
"loss": 0.0009,
"step": 14650
},
{
"epoch": 85.82,
"learning_rate": 2.7323076923076924e-06,
"loss": 0.0005,
"step": 14675
},
{
"epoch": 85.96,
"learning_rate": 2.7194871794871796e-06,
"loss": 0.0004,
"step": 14700
},
{
"epoch": 86.11,
"learning_rate": 2.706666666666667e-06,
"loss": 0.0004,
"step": 14725
},
{
"epoch": 86.26,
"learning_rate": 2.693846153846154e-06,
"loss": 0.0005,
"step": 14750
},
{
"epoch": 86.4,
"learning_rate": 2.6810256410256414e-06,
"loss": 0.0005,
"step": 14775
},
{
"epoch": 86.55,
"learning_rate": 2.6682051282051286e-06,
"loss": 0.0004,
"step": 14800
},
{
"epoch": 86.7,
"learning_rate": 2.655384615384616e-06,
"loss": 0.0004,
"step": 14825
},
{
"epoch": 86.84,
"learning_rate": 2.642564102564103e-06,
"loss": 0.0008,
"step": 14850
},
{
"epoch": 86.99,
"learning_rate": 2.6297435897435904e-06,
"loss": 0.0004,
"step": 14875
},
{
"epoch": 87.13,
"learning_rate": 2.6169230769230768e-06,
"loss": 0.0005,
"step": 14900
},
{
"epoch": 87.28,
"learning_rate": 2.604102564102564e-06,
"loss": 0.0003,
"step": 14925
},
{
"epoch": 87.43,
"learning_rate": 2.5912820512820513e-06,
"loss": 0.0003,
"step": 14950
},
{
"epoch": 87.57,
"learning_rate": 2.5784615384615385e-06,
"loss": 0.0006,
"step": 14975
},
{
"epoch": 87.72,
"learning_rate": 2.5656410256410257e-06,
"loss": 0.0006,
"step": 15000
},
{
"epoch": 87.72,
"eval_loss": 0.3880477547645569,
"eval_runtime": 1412.4459,
"eval_samples_per_second": 4.666,
"eval_steps_per_second": 0.292,
"eval_wer": 13.103225936448448,
"step": 15000
},
{
"epoch": 87.87,
"learning_rate": 2.552820512820513e-06,
"loss": 0.0004,
"step": 15025
},
{
"epoch": 88.01,
"learning_rate": 2.5400000000000002e-06,
"loss": 0.0004,
"step": 15050
},
{
"epoch": 88.16,
"learning_rate": 2.5271794871794875e-06,
"loss": 0.0006,
"step": 15075
},
{
"epoch": 88.3,
"learning_rate": 2.5143589743589747e-06,
"loss": 0.0007,
"step": 15100
},
{
"epoch": 88.45,
"learning_rate": 2.501538461538462e-06,
"loss": 0.0008,
"step": 15125
},
{
"epoch": 88.6,
"learning_rate": 2.488717948717949e-06,
"loss": 0.0007,
"step": 15150
},
{
"epoch": 88.74,
"learning_rate": 2.475897435897436e-06,
"loss": 0.0004,
"step": 15175
},
{
"epoch": 88.89,
"learning_rate": 2.4630769230769233e-06,
"loss": 0.0005,
"step": 15200
},
{
"epoch": 89.04,
"learning_rate": 2.4502564102564106e-06,
"loss": 0.0005,
"step": 15225
},
{
"epoch": 89.18,
"learning_rate": 2.437435897435898e-06,
"loss": 0.0005,
"step": 15250
},
{
"epoch": 89.33,
"learning_rate": 2.4246153846153846e-06,
"loss": 0.0006,
"step": 15275
},
{
"epoch": 89.47,
"learning_rate": 2.411794871794872e-06,
"loss": 0.0003,
"step": 15300
},
{
"epoch": 89.62,
"learning_rate": 2.398974358974359e-06,
"loss": 0.0003,
"step": 15325
},
{
"epoch": 89.77,
"learning_rate": 2.3861538461538464e-06,
"loss": 0.0004,
"step": 15350
},
{
"epoch": 89.91,
"learning_rate": 2.3733333333333336e-06,
"loss": 0.0005,
"step": 15375
},
{
"epoch": 90.06,
"learning_rate": 2.360512820512821e-06,
"loss": 0.0006,
"step": 15400
},
{
"epoch": 90.2,
"learning_rate": 2.3476923076923077e-06,
"loss": 0.0008,
"step": 15425
},
{
"epoch": 90.35,
"learning_rate": 2.334871794871795e-06,
"loss": 0.0006,
"step": 15450
},
{
"epoch": 90.5,
"learning_rate": 2.322051282051282e-06,
"loss": 0.0003,
"step": 15475
},
{
"epoch": 90.64,
"learning_rate": 2.3092307692307694e-06,
"loss": 0.0004,
"step": 15500
},
{
"epoch": 90.79,
"learning_rate": 2.2964102564102567e-06,
"loss": 0.0003,
"step": 15525
},
{
"epoch": 90.94,
"learning_rate": 2.283589743589744e-06,
"loss": 0.0004,
"step": 15550
},
{
"epoch": 91.08,
"learning_rate": 2.2707692307692308e-06,
"loss": 0.0004,
"step": 15575
},
{
"epoch": 91.23,
"learning_rate": 2.257948717948718e-06,
"loss": 0.0005,
"step": 15600
},
{
"epoch": 91.37,
"learning_rate": 2.2451282051282053e-06,
"loss": 0.0003,
"step": 15625
},
{
"epoch": 91.52,
"learning_rate": 2.2323076923076925e-06,
"loss": 0.0003,
"step": 15650
},
{
"epoch": 91.67,
"learning_rate": 2.2194871794871797e-06,
"loss": 0.0005,
"step": 15675
},
{
"epoch": 91.81,
"learning_rate": 2.206666666666667e-06,
"loss": 0.0003,
"step": 15700
},
{
"epoch": 91.96,
"learning_rate": 2.193846153846154e-06,
"loss": 0.0004,
"step": 15725
},
{
"epoch": 92.11,
"learning_rate": 2.181025641025641e-06,
"loss": 0.0002,
"step": 15750
},
{
"epoch": 92.25,
"learning_rate": 2.1682051282051283e-06,
"loss": 0.0003,
"step": 15775
},
{
"epoch": 92.4,
"learning_rate": 2.1553846153846156e-06,
"loss": 0.0003,
"step": 15800
},
{
"epoch": 92.54,
"learning_rate": 2.142564102564103e-06,
"loss": 0.0002,
"step": 15825
},
{
"epoch": 92.69,
"learning_rate": 2.12974358974359e-06,
"loss": 0.0002,
"step": 15850
},
{
"epoch": 92.84,
"learning_rate": 2.116923076923077e-06,
"loss": 0.0002,
"step": 15875
},
{
"epoch": 92.98,
"learning_rate": 2.104102564102564e-06,
"loss": 0.0002,
"step": 15900
},
{
"epoch": 93.13,
"learning_rate": 2.0912820512820514e-06,
"loss": 0.0002,
"step": 15925
},
{
"epoch": 93.27,
"learning_rate": 2.0784615384615386e-06,
"loss": 0.0002,
"step": 15950
},
{
"epoch": 93.42,
"learning_rate": 2.065641025641026e-06,
"loss": 0.0002,
"step": 15975
},
{
"epoch": 93.57,
"learning_rate": 2.052820512820513e-06,
"loss": 0.0002,
"step": 16000
},
{
"epoch": 93.57,
"eval_loss": 0.4004988372325897,
"eval_runtime": 1416.7381,
"eval_samples_per_second": 4.652,
"eval_steps_per_second": 0.291,
"eval_wer": 12.698220798323561,
"step": 16000
},
{
"epoch": 93.71,
"learning_rate": 2.04e-06,
"loss": 0.0002,
"step": 16025
},
{
"epoch": 93.86,
"learning_rate": 2.027179487179487e-06,
"loss": 0.0002,
"step": 16050
},
{
"epoch": 94.01,
"learning_rate": 2.0143589743589744e-06,
"loss": 0.0002,
"step": 16075
},
{
"epoch": 94.15,
"learning_rate": 2.0015384615384617e-06,
"loss": 0.0002,
"step": 16100
},
{
"epoch": 94.3,
"learning_rate": 1.988717948717949e-06,
"loss": 0.0003,
"step": 16125
},
{
"epoch": 94.44,
"learning_rate": 1.975897435897436e-06,
"loss": 0.0002,
"step": 16150
},
{
"epoch": 94.59,
"learning_rate": 1.963076923076923e-06,
"loss": 0.0002,
"step": 16175
},
{
"epoch": 94.74,
"learning_rate": 1.9502564102564103e-06,
"loss": 0.0003,
"step": 16200
},
{
"epoch": 94.88,
"learning_rate": 1.9374358974358975e-06,
"loss": 0.0003,
"step": 16225
},
{
"epoch": 95.03,
"learning_rate": 1.9246153846153848e-06,
"loss": 0.0003,
"step": 16250
},
{
"epoch": 95.18,
"learning_rate": 1.911794871794872e-06,
"loss": 0.0004,
"step": 16275
},
{
"epoch": 95.32,
"learning_rate": 1.8989743589743593e-06,
"loss": 0.0002,
"step": 16300
},
{
"epoch": 95.47,
"learning_rate": 1.8861538461538465e-06,
"loss": 0.0002,
"step": 16325
},
{
"epoch": 95.61,
"learning_rate": 1.8733333333333333e-06,
"loss": 0.0003,
"step": 16350
},
{
"epoch": 95.76,
"learning_rate": 1.8605128205128206e-06,
"loss": 0.0003,
"step": 16375
},
{
"epoch": 95.91,
"learning_rate": 1.8476923076923078e-06,
"loss": 0.0004,
"step": 16400
},
{
"epoch": 96.05,
"learning_rate": 1.834871794871795e-06,
"loss": 0.0003,
"step": 16425
},
{
"epoch": 96.2,
"learning_rate": 1.8220512820512823e-06,
"loss": 0.0002,
"step": 16450
},
{
"epoch": 96.35,
"learning_rate": 1.8092307692307696e-06,
"loss": 0.0003,
"step": 16475
},
{
"epoch": 96.49,
"learning_rate": 1.7964102564102564e-06,
"loss": 0.0003,
"step": 16500
},
{
"epoch": 96.64,
"learning_rate": 1.7835897435897436e-06,
"loss": 0.0002,
"step": 16525
},
{
"epoch": 96.78,
"learning_rate": 1.7707692307692309e-06,
"loss": 0.0003,
"step": 16550
},
{
"epoch": 96.93,
"learning_rate": 1.7579487179487181e-06,
"loss": 0.0002,
"step": 16575
},
{
"epoch": 97.08,
"learning_rate": 1.7451282051282054e-06,
"loss": 0.0004,
"step": 16600
},
{
"epoch": 97.22,
"learning_rate": 1.7323076923076926e-06,
"loss": 0.0003,
"step": 16625
},
{
"epoch": 97.37,
"learning_rate": 1.7194871794871795e-06,
"loss": 0.0003,
"step": 16650
},
{
"epoch": 97.51,
"learning_rate": 1.7066666666666667e-06,
"loss": 0.0002,
"step": 16675
},
{
"epoch": 97.66,
"learning_rate": 1.693846153846154e-06,
"loss": 0.0003,
"step": 16700
},
{
"epoch": 97.81,
"learning_rate": 1.6810256410256412e-06,
"loss": 0.0003,
"step": 16725
},
{
"epoch": 97.95,
"learning_rate": 1.6682051282051284e-06,
"loss": 0.0005,
"step": 16750
},
{
"epoch": 98.1,
"learning_rate": 1.6553846153846157e-06,
"loss": 0.0002,
"step": 16775
},
{
"epoch": 98.25,
"learning_rate": 1.6425641025641025e-06,
"loss": 0.0002,
"step": 16800
},
{
"epoch": 98.39,
"learning_rate": 1.6297435897435898e-06,
"loss": 0.0002,
"step": 16825
},
{
"epoch": 98.54,
"learning_rate": 1.616923076923077e-06,
"loss": 0.0002,
"step": 16850
},
{
"epoch": 98.68,
"learning_rate": 1.6041025641025643e-06,
"loss": 0.0002,
"step": 16875
},
{
"epoch": 98.83,
"learning_rate": 1.5912820512820515e-06,
"loss": 0.0002,
"step": 16900
},
{
"epoch": 98.98,
"learning_rate": 1.5784615384615388e-06,
"loss": 0.0002,
"step": 16925
},
{
"epoch": 99.12,
"learning_rate": 1.5656410256410256e-06,
"loss": 0.0002,
"step": 16950
},
{
"epoch": 99.27,
"learning_rate": 1.5528205128205128e-06,
"loss": 0.0002,
"step": 16975
},
{
"epoch": 99.42,
"learning_rate": 1.54e-06,
"loss": 0.0002,
"step": 17000
},
{
"epoch": 99.42,
"eval_loss": 0.4003838002681732,
"eval_runtime": 1422.2846,
"eval_samples_per_second": 4.634,
"eval_steps_per_second": 0.29,
"eval_wer": 13.15158475891112,
"step": 17000
},
{
"epoch": 99.56,
"learning_rate": 1.5271794871794873e-06,
"loss": 0.0002,
"step": 17025
},
{
"epoch": 99.71,
"learning_rate": 1.5143589743589746e-06,
"loss": 0.0001,
"step": 17050
},
{
"epoch": 99.85,
"learning_rate": 1.5015384615384618e-06,
"loss": 0.0002,
"step": 17075
},
{
"epoch": 100.0,
"learning_rate": 1.4887179487179486e-06,
"loss": 0.0001,
"step": 17100
},
{
"epoch": 100.15,
"learning_rate": 1.475897435897436e-06,
"loss": 0.0001,
"step": 17125
},
{
"epoch": 100.29,
"learning_rate": 1.4630769230769231e-06,
"loss": 0.0001,
"step": 17150
},
{
"epoch": 100.44,
"learning_rate": 1.4502564102564104e-06,
"loss": 0.0002,
"step": 17175
},
{
"epoch": 100.58,
"learning_rate": 1.4374358974358976e-06,
"loss": 0.0002,
"step": 17200
},
{
"epoch": 100.73,
"learning_rate": 1.4246153846153849e-06,
"loss": 0.0002,
"step": 17225
},
{
"epoch": 100.88,
"learning_rate": 1.4117948717948717e-06,
"loss": 0.0003,
"step": 17250
},
{
"epoch": 101.02,
"learning_rate": 1.398974358974359e-06,
"loss": 0.0003,
"step": 17275
},
{
"epoch": 101.17,
"learning_rate": 1.3861538461538462e-06,
"loss": 0.0002,
"step": 17300
},
{
"epoch": 101.32,
"learning_rate": 1.3733333333333335e-06,
"loss": 0.0002,
"step": 17325
},
{
"epoch": 101.46,
"learning_rate": 1.3605128205128207e-06,
"loss": 0.0001,
"step": 17350
},
{
"epoch": 101.61,
"learning_rate": 1.347692307692308e-06,
"loss": 0.0002,
"step": 17375
},
{
"epoch": 101.75,
"learning_rate": 1.3348717948717948e-06,
"loss": 0.0001,
"step": 17400
},
{
"epoch": 101.9,
"learning_rate": 1.322051282051282e-06,
"loss": 0.0002,
"step": 17425
},
{
"epoch": 102.05,
"learning_rate": 1.3092307692307693e-06,
"loss": 0.0002,
"step": 17450
},
{
"epoch": 102.19,
"learning_rate": 1.2964102564102565e-06,
"loss": 0.0002,
"step": 17475
},
{
"epoch": 102.34,
"learning_rate": 1.2835897435897438e-06,
"loss": 0.0001,
"step": 17500
},
{
"epoch": 102.49,
"learning_rate": 1.270769230769231e-06,
"loss": 0.0002,
"step": 17525
},
{
"epoch": 102.63,
"learning_rate": 1.2579487179487178e-06,
"loss": 0.0001,
"step": 17550
},
{
"epoch": 102.78,
"learning_rate": 1.2451282051282053e-06,
"loss": 0.0001,
"step": 17575
},
{
"epoch": 102.92,
"learning_rate": 1.2323076923076923e-06,
"loss": 0.0002,
"step": 17600
},
{
"epoch": 103.07,
"learning_rate": 1.2194871794871796e-06,
"loss": 0.0002,
"step": 17625
},
{
"epoch": 103.22,
"learning_rate": 1.2066666666666668e-06,
"loss": 0.0001,
"step": 17650
},
{
"epoch": 103.36,
"learning_rate": 1.1938461538461539e-06,
"loss": 0.0002,
"step": 17675
},
{
"epoch": 103.51,
"learning_rate": 1.1810256410256411e-06,
"loss": 0.0002,
"step": 17700
},
{
"epoch": 103.65,
"learning_rate": 1.1682051282051284e-06,
"loss": 0.0001,
"step": 17725
},
{
"epoch": 103.8,
"learning_rate": 1.1553846153846154e-06,
"loss": 0.0001,
"step": 17750
},
{
"epoch": 103.95,
"learning_rate": 1.1425641025641026e-06,
"loss": 0.0001,
"step": 17775
},
{
"epoch": 104.09,
"learning_rate": 1.12974358974359e-06,
"loss": 0.0002,
"step": 17800
},
{
"epoch": 104.24,
"learning_rate": 1.116923076923077e-06,
"loss": 0.0002,
"step": 17825
},
{
"epoch": 104.39,
"learning_rate": 1.1041025641025642e-06,
"loss": 0.0001,
"step": 17850
},
{
"epoch": 104.53,
"learning_rate": 1.0912820512820514e-06,
"loss": 0.0001,
"step": 17875
},
{
"epoch": 104.68,
"learning_rate": 1.0784615384615385e-06,
"loss": 0.0001,
"step": 17900
},
{
"epoch": 104.82,
"learning_rate": 1.0656410256410257e-06,
"loss": 0.0001,
"step": 17925
},
{
"epoch": 104.97,
"learning_rate": 1.052820512820513e-06,
"loss": 0.0002,
"step": 17950
},
{
"epoch": 105.12,
"learning_rate": 1.04e-06,
"loss": 0.0002,
"step": 17975
},
{
"epoch": 105.26,
"learning_rate": 1.0271794871794872e-06,
"loss": 0.0001,
"step": 18000
},
{
"epoch": 105.26,
"eval_loss": 0.4139898419380188,
"eval_runtime": 1412.6561,
"eval_samples_per_second": 4.666,
"eval_steps_per_second": 0.292,
"eval_wer": 12.87352152975075,
"step": 18000
},
{
"epoch": 105.41,
"learning_rate": 1.0143589743589745e-06,
"loss": 0.0001,
"step": 18025
},
{
"epoch": 105.56,
"learning_rate": 1.0015384615384615e-06,
"loss": 0.0001,
"step": 18050
},
{
"epoch": 105.7,
"learning_rate": 9.887179487179488e-07,
"loss": 0.0001,
"step": 18075
},
{
"epoch": 105.85,
"learning_rate": 9.75897435897436e-07,
"loss": 0.0001,
"step": 18100
},
{
"epoch": 105.99,
"learning_rate": 9.63076923076923e-07,
"loss": 0.0001,
"step": 18125
},
{
"epoch": 106.14,
"learning_rate": 9.502564102564103e-07,
"loss": 0.0002,
"step": 18150
},
{
"epoch": 106.29,
"learning_rate": 9.374358974358976e-07,
"loss": 0.0002,
"step": 18175
},
{
"epoch": 106.43,
"learning_rate": 9.246153846153846e-07,
"loss": 0.0001,
"step": 18200
},
{
"epoch": 106.58,
"learning_rate": 9.117948717948718e-07,
"loss": 0.0002,
"step": 18225
},
{
"epoch": 106.73,
"learning_rate": 8.989743589743591e-07,
"loss": 0.0002,
"step": 18250
},
{
"epoch": 106.87,
"learning_rate": 8.861538461538461e-07,
"loss": 0.0001,
"step": 18275
},
{
"epoch": 107.02,
"learning_rate": 8.733333333333334e-07,
"loss": 0.0001,
"step": 18300
},
{
"epoch": 107.16,
"learning_rate": 8.605128205128206e-07,
"loss": 0.0001,
"step": 18325
},
{
"epoch": 107.31,
"learning_rate": 8.476923076923077e-07,
"loss": 0.0002,
"step": 18350
},
{
"epoch": 107.46,
"learning_rate": 8.348717948717949e-07,
"loss": 0.0001,
"step": 18375
},
{
"epoch": 107.6,
"learning_rate": 8.220512820512822e-07,
"loss": 0.0001,
"step": 18400
},
{
"epoch": 107.75,
"learning_rate": 8.092307692307692e-07,
"loss": 0.0002,
"step": 18425
},
{
"epoch": 107.89,
"learning_rate": 7.964102564102564e-07,
"loss": 0.0002,
"step": 18450
},
{
"epoch": 108.04,
"learning_rate": 7.835897435897437e-07,
"loss": 0.0002,
"step": 18475
},
{
"epoch": 108.19,
"learning_rate": 7.707692307692307e-07,
"loss": 0.0001,
"step": 18500
},
{
"epoch": 108.33,
"learning_rate": 7.57948717948718e-07,
"loss": 0.0001,
"step": 18525
},
{
"epoch": 108.48,
"learning_rate": 7.451282051282052e-07,
"loss": 0.0001,
"step": 18550
},
{
"epoch": 108.63,
"learning_rate": 7.323076923076923e-07,
"loss": 0.0001,
"step": 18575
},
{
"epoch": 108.77,
"learning_rate": 7.194871794871795e-07,
"loss": 0.0001,
"step": 18600
},
{
"epoch": 108.92,
"learning_rate": 7.066666666666667e-07,
"loss": 0.0002,
"step": 18625
},
{
"epoch": 109.06,
"learning_rate": 6.938461538461538e-07,
"loss": 0.0001,
"step": 18650
},
{
"epoch": 109.21,
"learning_rate": 6.81025641025641e-07,
"loss": 0.0002,
"step": 18675
},
{
"epoch": 109.36,
"learning_rate": 6.682051282051283e-07,
"loss": 0.0001,
"step": 18700
},
{
"epoch": 109.5,
"learning_rate": 6.553846153846154e-07,
"loss": 0.0001,
"step": 18725
},
{
"epoch": 109.65,
"learning_rate": 6.425641025641026e-07,
"loss": 0.0001,
"step": 18750
},
{
"epoch": 109.8,
"learning_rate": 6.297435897435898e-07,
"loss": 0.0002,
"step": 18775
},
{
"epoch": 109.94,
"learning_rate": 6.16923076923077e-07,
"loss": 0.0001,
"step": 18800
},
{
"epoch": 110.09,
"learning_rate": 6.041025641025641e-07,
"loss": 0.0002,
"step": 18825
},
{
"epoch": 110.23,
"learning_rate": 5.912820512820513e-07,
"loss": 0.0001,
"step": 18850
},
{
"epoch": 110.38,
"learning_rate": 5.784615384615385e-07,
"loss": 0.0001,
"step": 18875
},
{
"epoch": 110.53,
"learning_rate": 5.656410256410256e-07,
"loss": 0.0001,
"step": 18900
},
{
"epoch": 110.67,
"learning_rate": 5.528205128205129e-07,
"loss": 0.0003,
"step": 18925
},
{
"epoch": 110.82,
"learning_rate": 5.4e-07,
"loss": 0.0001,
"step": 18950
},
{
"epoch": 110.96,
"learning_rate": 5.271794871794872e-07,
"loss": 0.0001,
"step": 18975
},
{
"epoch": 111.11,
"learning_rate": 5.143589743589744e-07,
"loss": 0.0001,
"step": 19000
},
{
"epoch": 111.11,
"eval_loss": 0.41314777731895447,
"eval_runtime": 1411.4935,
"eval_samples_per_second": 4.67,
"eval_steps_per_second": 0.292,
"eval_wer": 12.512845312216648,
"step": 19000
},
{
"epoch": 111.26,
"learning_rate": 5.015384615384615e-07,
"loss": 0.0001,
"step": 19025
},
{
"epoch": 111.4,
"learning_rate": 4.887179487179487e-07,
"loss": 0.0001,
"step": 19050
},
{
"epoch": 111.55,
"learning_rate": 4.7589743589743594e-07,
"loss": 0.0001,
"step": 19075
},
{
"epoch": 111.7,
"learning_rate": 4.630769230769231e-07,
"loss": 0.0001,
"step": 19100
},
{
"epoch": 111.84,
"learning_rate": 4.502564102564103e-07,
"loss": 0.0001,
"step": 19125
},
{
"epoch": 111.99,
"learning_rate": 4.3743589743589747e-07,
"loss": 0.0001,
"step": 19150
},
{
"epoch": 112.13,
"learning_rate": 4.246153846153846e-07,
"loss": 0.0001,
"step": 19175
},
{
"epoch": 112.28,
"learning_rate": 4.1179487179487186e-07,
"loss": 0.0002,
"step": 19200
},
{
"epoch": 112.43,
"learning_rate": 3.98974358974359e-07,
"loss": 0.0001,
"step": 19225
},
{
"epoch": 112.57,
"learning_rate": 3.8615384615384615e-07,
"loss": 0.0001,
"step": 19250
},
{
"epoch": 112.72,
"learning_rate": 3.733333333333334e-07,
"loss": 0.0001,
"step": 19275
},
{
"epoch": 112.87,
"learning_rate": 3.6051282051282054e-07,
"loss": 0.0002,
"step": 19300
},
{
"epoch": 113.01,
"learning_rate": 3.476923076923077e-07,
"loss": 0.0001,
"step": 19325
},
{
"epoch": 113.16,
"learning_rate": 3.348717948717949e-07,
"loss": 0.0001,
"step": 19350
},
{
"epoch": 113.3,
"learning_rate": 3.2205128205128207e-07,
"loss": 0.0002,
"step": 19375
},
{
"epoch": 113.45,
"learning_rate": 3.0923076923076926e-07,
"loss": 0.0001,
"step": 19400
},
{
"epoch": 113.6,
"learning_rate": 2.964102564102564e-07,
"loss": 0.0001,
"step": 19425
},
{
"epoch": 113.74,
"learning_rate": 2.835897435897436e-07,
"loss": 0.0002,
"step": 19450
},
{
"epoch": 113.89,
"learning_rate": 2.707692307692308e-07,
"loss": 0.0001,
"step": 19475
},
{
"epoch": 114.04,
"learning_rate": 2.5794871794871794e-07,
"loss": 0.0001,
"step": 19500
},
{
"epoch": 114.18,
"learning_rate": 2.4512820512820513e-07,
"loss": 0.0002,
"step": 19525
},
{
"epoch": 114.33,
"learning_rate": 2.3230769230769233e-07,
"loss": 0.0001,
"step": 19550
},
{
"epoch": 114.47,
"learning_rate": 2.194871794871795e-07,
"loss": 0.0001,
"step": 19575
},
{
"epoch": 114.62,
"learning_rate": 2.066666666666667e-07,
"loss": 0.0002,
"step": 19600
},
{
"epoch": 114.77,
"learning_rate": 1.9384615384615386e-07,
"loss": 0.0001,
"step": 19625
},
{
"epoch": 114.91,
"learning_rate": 1.8102564102564103e-07,
"loss": 0.0001,
"step": 19650
},
{
"epoch": 115.06,
"learning_rate": 1.6820512820512822e-07,
"loss": 0.0001,
"step": 19675
},
{
"epoch": 115.2,
"learning_rate": 1.553846153846154e-07,
"loss": 0.0001,
"step": 19700
},
{
"epoch": 115.35,
"learning_rate": 1.4256410256410259e-07,
"loss": 0.0002,
"step": 19725
},
{
"epoch": 115.5,
"learning_rate": 1.2974358974358975e-07,
"loss": 0.0001,
"step": 19750
},
{
"epoch": 115.64,
"learning_rate": 1.1692307692307694e-07,
"loss": 0.0002,
"step": 19775
},
{
"epoch": 115.79,
"learning_rate": 1.041025641025641e-07,
"loss": 0.0001,
"step": 19800
},
{
"epoch": 115.94,
"learning_rate": 9.128205128205129e-08,
"loss": 0.0001,
"step": 19825
},
{
"epoch": 116.08,
"learning_rate": 7.846153846153847e-08,
"loss": 0.0002,
"step": 19850
},
{
"epoch": 116.23,
"learning_rate": 6.564102564102564e-08,
"loss": 0.0001,
"step": 19875
},
{
"epoch": 116.37,
"learning_rate": 5.2820512820512826e-08,
"loss": 0.0001,
"step": 19900
},
{
"epoch": 116.52,
"learning_rate": 4e-08,
"loss": 0.0001,
"step": 19925
},
{
"epoch": 116.67,
"learning_rate": 2.717948717948718e-08,
"loss": 0.0001,
"step": 19950
},
{
"epoch": 116.81,
"learning_rate": 1.4358974358974361e-08,
"loss": 0.0001,
"step": 19975
},
{
"epoch": 116.96,
"learning_rate": 1.5384615384615387e-09,
"loss": 0.0001,
"step": 20000
},
{
"epoch": 116.96,
"eval_loss": 0.41206619143486023,
"eval_runtime": 1417.0237,
"eval_samples_per_second": 4.651,
"eval_steps_per_second": 0.291,
"eval_wer": 12.627697515565494,
"step": 20000
},
{
"epoch": 116.96,
"step": 20000,
"total_flos": 2.711255422825263e+21,
"train_loss": 0.028188849427364767,
"train_runtime": 142803.5202,
"train_samples_per_second": 8.963,
"train_steps_per_second": 0.14
}
],
"logging_steps": 25,
"max_steps": 20000,
"num_input_tokens_seen": 0,
"num_train_epochs": 117,
"save_steps": 1000,
"total_flos": 2.711255422825263e+21,
"train_batch_size": 32,
"trial_name": null,
"trial_params": null
}