whisper-large-gl-train / trainer_state.json
zuazo's picture
End of training
741bdce verified
{
"best_metric": 6.2705357451665344,
"best_model_checkpoint": "./checkpoint-2000",
"epoch": 116.61807580174927,
"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": 4.6000000000000004e-07,
"loss": 2.654,
"step": 25
},
{
"epoch": 0.29,
"learning_rate": 9.600000000000001e-07,
"loss": 1.5215,
"step": 50
},
{
"epoch": 0.44,
"learning_rate": 1.44e-06,
"loss": 0.8797,
"step": 75
},
{
"epoch": 0.58,
"learning_rate": 1.94e-06,
"loss": 0.6317,
"step": 100
},
{
"epoch": 0.73,
"learning_rate": 2.4400000000000004e-06,
"loss": 0.5446,
"step": 125
},
{
"epoch": 0.87,
"learning_rate": 2.9400000000000002e-06,
"loss": 0.4752,
"step": 150
},
{
"epoch": 1.02,
"learning_rate": 3.4200000000000007e-06,
"loss": 0.4259,
"step": 175
},
{
"epoch": 1.17,
"learning_rate": 3.920000000000001e-06,
"loss": 0.3558,
"step": 200
},
{
"epoch": 1.31,
"learning_rate": 4.42e-06,
"loss": 0.3419,
"step": 225
},
{
"epoch": 1.46,
"learning_rate": 4.92e-06,
"loss": 0.3128,
"step": 250
},
{
"epoch": 1.6,
"learning_rate": 5.420000000000001e-06,
"loss": 0.2941,
"step": 275
},
{
"epoch": 1.75,
"learning_rate": 5.92e-06,
"loss": 0.2744,
"step": 300
},
{
"epoch": 1.9,
"learning_rate": 6.42e-06,
"loss": 0.2692,
"step": 325
},
{
"epoch": 2.04,
"learning_rate": 6.92e-06,
"loss": 0.2273,
"step": 350
},
{
"epoch": 2.19,
"learning_rate": 7.420000000000001e-06,
"loss": 0.1843,
"step": 375
},
{
"epoch": 2.33,
"learning_rate": 7.92e-06,
"loss": 0.1862,
"step": 400
},
{
"epoch": 2.48,
"learning_rate": 8.42e-06,
"loss": 0.1809,
"step": 425
},
{
"epoch": 2.62,
"learning_rate": 8.920000000000001e-06,
"loss": 0.1849,
"step": 450
},
{
"epoch": 2.77,
"learning_rate": 9.42e-06,
"loss": 0.1824,
"step": 475
},
{
"epoch": 2.92,
"learning_rate": 9.920000000000002e-06,
"loss": 0.1849,
"step": 500
},
{
"epoch": 3.06,
"learning_rate": 9.98923076923077e-06,
"loss": 0.1391,
"step": 525
},
{
"epoch": 3.21,
"learning_rate": 9.976410256410257e-06,
"loss": 0.1147,
"step": 550
},
{
"epoch": 3.35,
"learning_rate": 9.963589743589744e-06,
"loss": 0.1103,
"step": 575
},
{
"epoch": 3.5,
"learning_rate": 9.950769230769232e-06,
"loss": 0.1233,
"step": 600
},
{
"epoch": 3.64,
"learning_rate": 9.937948717948719e-06,
"loss": 0.1124,
"step": 625
},
{
"epoch": 3.79,
"learning_rate": 9.925128205128206e-06,
"loss": 0.1136,
"step": 650
},
{
"epoch": 3.94,
"learning_rate": 9.912307692307693e-06,
"loss": 0.1111,
"step": 675
},
{
"epoch": 4.08,
"learning_rate": 9.899487179487181e-06,
"loss": 0.0889,
"step": 700
},
{
"epoch": 4.23,
"learning_rate": 9.886666666666668e-06,
"loss": 0.0672,
"step": 725
},
{
"epoch": 4.37,
"learning_rate": 9.873846153846155e-06,
"loss": 0.0714,
"step": 750
},
{
"epoch": 4.52,
"learning_rate": 9.861025641025642e-06,
"loss": 0.069,
"step": 775
},
{
"epoch": 4.66,
"learning_rate": 9.84820512820513e-06,
"loss": 0.072,
"step": 800
},
{
"epoch": 4.81,
"learning_rate": 9.835384615384617e-06,
"loss": 0.0708,
"step": 825
},
{
"epoch": 4.96,
"learning_rate": 9.822564102564104e-06,
"loss": 0.0725,
"step": 850
},
{
"epoch": 5.1,
"learning_rate": 9.80974358974359e-06,
"loss": 0.0546,
"step": 875
},
{
"epoch": 5.25,
"learning_rate": 9.796923076923077e-06,
"loss": 0.0451,
"step": 900
},
{
"epoch": 5.39,
"learning_rate": 9.784102564102564e-06,
"loss": 0.0482,
"step": 925
},
{
"epoch": 5.54,
"learning_rate": 9.771282051282051e-06,
"loss": 0.0454,
"step": 950
},
{
"epoch": 5.69,
"learning_rate": 9.75846153846154e-06,
"loss": 0.0464,
"step": 975
},
{
"epoch": 5.83,
"learning_rate": 9.745641025641026e-06,
"loss": 0.0476,
"step": 1000
},
{
"epoch": 5.83,
"eval_loss": 0.18290099501609802,
"eval_runtime": 1318.3045,
"eval_samples_per_second": 4.852,
"eval_steps_per_second": 0.303,
"eval_wer": 6.35354338878705,
"step": 1000
},
{
"epoch": 5.98,
"learning_rate": 9.732820512820513e-06,
"loss": 0.0485,
"step": 1025
},
{
"epoch": 6.12,
"learning_rate": 9.72e-06,
"loss": 0.0332,
"step": 1050
},
{
"epoch": 6.27,
"learning_rate": 9.707179487179487e-06,
"loss": 0.0286,
"step": 1075
},
{
"epoch": 6.41,
"learning_rate": 9.694358974358975e-06,
"loss": 0.0307,
"step": 1100
},
{
"epoch": 6.56,
"learning_rate": 9.681538461538462e-06,
"loss": 0.0324,
"step": 1125
},
{
"epoch": 6.71,
"learning_rate": 9.668717948717949e-06,
"loss": 0.0316,
"step": 1150
},
{
"epoch": 6.85,
"learning_rate": 9.655897435897436e-06,
"loss": 0.0326,
"step": 1175
},
{
"epoch": 7.0,
"learning_rate": 9.643076923076924e-06,
"loss": 0.0349,
"step": 1200
},
{
"epoch": 7.14,
"learning_rate": 9.630256410256411e-06,
"loss": 0.0184,
"step": 1225
},
{
"epoch": 7.29,
"learning_rate": 9.617435897435898e-06,
"loss": 0.0234,
"step": 1250
},
{
"epoch": 7.43,
"learning_rate": 9.604615384615385e-06,
"loss": 0.0221,
"step": 1275
},
{
"epoch": 7.58,
"learning_rate": 9.591794871794873e-06,
"loss": 0.0242,
"step": 1300
},
{
"epoch": 7.73,
"learning_rate": 9.57897435897436e-06,
"loss": 0.0243,
"step": 1325
},
{
"epoch": 7.87,
"learning_rate": 9.566153846153847e-06,
"loss": 0.023,
"step": 1350
},
{
"epoch": 8.02,
"learning_rate": 9.553333333333334e-06,
"loss": 0.0252,
"step": 1375
},
{
"epoch": 8.16,
"learning_rate": 9.540512820512822e-06,
"loss": 0.0148,
"step": 1400
},
{
"epoch": 8.31,
"learning_rate": 9.52769230769231e-06,
"loss": 0.0197,
"step": 1425
},
{
"epoch": 8.45,
"learning_rate": 9.514871794871796e-06,
"loss": 0.0183,
"step": 1450
},
{
"epoch": 8.6,
"learning_rate": 9.502051282051283e-06,
"loss": 0.02,
"step": 1475
},
{
"epoch": 8.75,
"learning_rate": 9.48923076923077e-06,
"loss": 0.0166,
"step": 1500
},
{
"epoch": 8.89,
"learning_rate": 9.476410256410257e-06,
"loss": 0.0189,
"step": 1525
},
{
"epoch": 9.04,
"learning_rate": 9.463589743589743e-06,
"loss": 0.0175,
"step": 1550
},
{
"epoch": 9.18,
"learning_rate": 9.450769230769232e-06,
"loss": 0.0124,
"step": 1575
},
{
"epoch": 9.33,
"learning_rate": 9.437948717948719e-06,
"loss": 0.0122,
"step": 1600
},
{
"epoch": 9.48,
"learning_rate": 9.425128205128206e-06,
"loss": 0.012,
"step": 1625
},
{
"epoch": 9.62,
"learning_rate": 9.412307692307692e-06,
"loss": 0.0127,
"step": 1650
},
{
"epoch": 9.77,
"learning_rate": 9.39948717948718e-06,
"loss": 0.0131,
"step": 1675
},
{
"epoch": 9.91,
"learning_rate": 9.386666666666668e-06,
"loss": 0.0147,
"step": 1700
},
{
"epoch": 10.06,
"learning_rate": 9.373846153846155e-06,
"loss": 0.013,
"step": 1725
},
{
"epoch": 10.2,
"learning_rate": 9.361025641025641e-06,
"loss": 0.01,
"step": 1750
},
{
"epoch": 10.35,
"learning_rate": 9.348205128205128e-06,
"loss": 0.0094,
"step": 1775
},
{
"epoch": 10.5,
"learning_rate": 9.335384615384617e-06,
"loss": 0.0119,
"step": 1800
},
{
"epoch": 10.64,
"learning_rate": 9.322564102564104e-06,
"loss": 0.0102,
"step": 1825
},
{
"epoch": 10.79,
"learning_rate": 9.30974358974359e-06,
"loss": 0.0109,
"step": 1850
},
{
"epoch": 10.93,
"learning_rate": 9.296923076923077e-06,
"loss": 0.0101,
"step": 1875
},
{
"epoch": 11.08,
"learning_rate": 9.284102564102566e-06,
"loss": 0.0084,
"step": 1900
},
{
"epoch": 11.22,
"learning_rate": 9.271282051282053e-06,
"loss": 0.008,
"step": 1925
},
{
"epoch": 11.37,
"learning_rate": 9.25846153846154e-06,
"loss": 0.0084,
"step": 1950
},
{
"epoch": 11.52,
"learning_rate": 9.245641025641026e-06,
"loss": 0.009,
"step": 1975
},
{
"epoch": 11.66,
"learning_rate": 9.232820512820515e-06,
"loss": 0.008,
"step": 2000
},
{
"epoch": 11.66,
"eval_loss": 0.2223636507987976,
"eval_runtime": 1315.8355,
"eval_samples_per_second": 4.862,
"eval_steps_per_second": 0.304,
"eval_wer": 6.2705357451665344,
"step": 2000
},
{
"epoch": 11.81,
"learning_rate": 9.220000000000002e-06,
"loss": 0.0106,
"step": 2025
},
{
"epoch": 11.95,
"learning_rate": 9.207179487179488e-06,
"loss": 0.0081,
"step": 2050
},
{
"epoch": 12.1,
"learning_rate": 9.194358974358975e-06,
"loss": 0.0072,
"step": 2075
},
{
"epoch": 12.24,
"learning_rate": 9.181538461538464e-06,
"loss": 0.0085,
"step": 2100
},
{
"epoch": 12.39,
"learning_rate": 9.168717948717949e-06,
"loss": 0.0069,
"step": 2125
},
{
"epoch": 12.54,
"learning_rate": 9.155897435897436e-06,
"loss": 0.0069,
"step": 2150
},
{
"epoch": 12.68,
"learning_rate": 9.143076923076924e-06,
"loss": 0.0069,
"step": 2175
},
{
"epoch": 12.83,
"learning_rate": 9.130256410256411e-06,
"loss": 0.008,
"step": 2200
},
{
"epoch": 12.97,
"learning_rate": 9.117435897435898e-06,
"loss": 0.0067,
"step": 2225
},
{
"epoch": 13.12,
"learning_rate": 9.104615384615385e-06,
"loss": 0.0058,
"step": 2250
},
{
"epoch": 13.27,
"learning_rate": 9.091794871794873e-06,
"loss": 0.007,
"step": 2275
},
{
"epoch": 13.41,
"learning_rate": 9.07897435897436e-06,
"loss": 0.0063,
"step": 2300
},
{
"epoch": 13.56,
"learning_rate": 9.066153846153847e-06,
"loss": 0.0066,
"step": 2325
},
{
"epoch": 13.7,
"learning_rate": 9.053333333333334e-06,
"loss": 0.007,
"step": 2350
},
{
"epoch": 13.85,
"learning_rate": 9.04051282051282e-06,
"loss": 0.0083,
"step": 2375
},
{
"epoch": 13.99,
"learning_rate": 9.027692307692309e-06,
"loss": 0.0066,
"step": 2400
},
{
"epoch": 14.14,
"learning_rate": 9.014871794871796e-06,
"loss": 0.0064,
"step": 2425
},
{
"epoch": 14.29,
"learning_rate": 9.002051282051283e-06,
"loss": 0.0056,
"step": 2450
},
{
"epoch": 14.43,
"learning_rate": 8.98923076923077e-06,
"loss": 0.0055,
"step": 2475
},
{
"epoch": 14.58,
"learning_rate": 8.976410256410258e-06,
"loss": 0.0052,
"step": 2500
},
{
"epoch": 14.72,
"learning_rate": 8.963589743589745e-06,
"loss": 0.0049,
"step": 2525
},
{
"epoch": 14.87,
"learning_rate": 8.950769230769232e-06,
"loss": 0.0084,
"step": 2550
},
{
"epoch": 15.01,
"learning_rate": 8.937948717948718e-06,
"loss": 0.0071,
"step": 2575
},
{
"epoch": 15.16,
"learning_rate": 8.925128205128207e-06,
"loss": 0.0048,
"step": 2600
},
{
"epoch": 15.31,
"learning_rate": 8.912307692307694e-06,
"loss": 0.0051,
"step": 2625
},
{
"epoch": 15.45,
"learning_rate": 8.89948717948718e-06,
"loss": 0.0049,
"step": 2650
},
{
"epoch": 15.6,
"learning_rate": 8.886666666666667e-06,
"loss": 0.0056,
"step": 2675
},
{
"epoch": 15.74,
"learning_rate": 8.873846153846156e-06,
"loss": 0.0063,
"step": 2700
},
{
"epoch": 15.89,
"learning_rate": 8.861025641025641e-06,
"loss": 0.0062,
"step": 2725
},
{
"epoch": 16.03,
"learning_rate": 8.848205128205128e-06,
"loss": 0.0072,
"step": 2750
},
{
"epoch": 16.18,
"learning_rate": 8.835384615384616e-06,
"loss": 0.005,
"step": 2775
},
{
"epoch": 16.33,
"learning_rate": 8.822564102564103e-06,
"loss": 0.0048,
"step": 2800
},
{
"epoch": 16.47,
"learning_rate": 8.80974358974359e-06,
"loss": 0.0041,
"step": 2825
},
{
"epoch": 16.62,
"learning_rate": 8.796923076923077e-06,
"loss": 0.0045,
"step": 2850
},
{
"epoch": 16.76,
"learning_rate": 8.784102564102565e-06,
"loss": 0.0052,
"step": 2875
},
{
"epoch": 16.91,
"learning_rate": 8.771282051282052e-06,
"loss": 0.0052,
"step": 2900
},
{
"epoch": 17.06,
"learning_rate": 8.758461538461539e-06,
"loss": 0.0053,
"step": 2925
},
{
"epoch": 17.2,
"learning_rate": 8.745641025641026e-06,
"loss": 0.0054,
"step": 2950
},
{
"epoch": 17.35,
"learning_rate": 8.732820512820513e-06,
"loss": 0.0051,
"step": 2975
},
{
"epoch": 17.49,
"learning_rate": 8.720000000000001e-06,
"loss": 0.0043,
"step": 3000
},
{
"epoch": 17.49,
"eval_loss": 0.23600055277347565,
"eval_runtime": 1316.0686,
"eval_samples_per_second": 4.861,
"eval_steps_per_second": 0.304,
"eval_wer": 6.339708781516965,
"step": 3000
},
{
"epoch": 17.64,
"learning_rate": 8.707179487179488e-06,
"loss": 0.0038,
"step": 3025
},
{
"epoch": 17.78,
"learning_rate": 8.694358974358975e-06,
"loss": 0.0058,
"step": 3050
},
{
"epoch": 17.93,
"learning_rate": 8.681538461538462e-06,
"loss": 0.0038,
"step": 3075
},
{
"epoch": 18.08,
"learning_rate": 8.66871794871795e-06,
"loss": 0.0041,
"step": 3100
},
{
"epoch": 18.22,
"learning_rate": 8.655897435897437e-06,
"loss": 0.0047,
"step": 3125
},
{
"epoch": 18.37,
"learning_rate": 8.643076923076924e-06,
"loss": 0.0036,
"step": 3150
},
{
"epoch": 18.51,
"learning_rate": 8.63025641025641e-06,
"loss": 0.004,
"step": 3175
},
{
"epoch": 18.66,
"learning_rate": 8.6174358974359e-06,
"loss": 0.0044,
"step": 3200
},
{
"epoch": 18.8,
"learning_rate": 8.604615384615386e-06,
"loss": 0.0036,
"step": 3225
},
{
"epoch": 18.95,
"learning_rate": 8.591794871794873e-06,
"loss": 0.0048,
"step": 3250
},
{
"epoch": 19.1,
"learning_rate": 8.57897435897436e-06,
"loss": 0.0034,
"step": 3275
},
{
"epoch": 19.24,
"learning_rate": 8.566153846153848e-06,
"loss": 0.0047,
"step": 3300
},
{
"epoch": 19.39,
"learning_rate": 8.553333333333333e-06,
"loss": 0.0043,
"step": 3325
},
{
"epoch": 19.53,
"learning_rate": 8.54051282051282e-06,
"loss": 0.0043,
"step": 3350
},
{
"epoch": 19.68,
"learning_rate": 8.527692307692309e-06,
"loss": 0.0035,
"step": 3375
},
{
"epoch": 19.83,
"learning_rate": 8.514871794871795e-06,
"loss": 0.0034,
"step": 3400
},
{
"epoch": 19.97,
"learning_rate": 8.502051282051282e-06,
"loss": 0.0047,
"step": 3425
},
{
"epoch": 20.12,
"learning_rate": 8.489230769230769e-06,
"loss": 0.0027,
"step": 3450
},
{
"epoch": 20.26,
"learning_rate": 8.476410256410258e-06,
"loss": 0.0033,
"step": 3475
},
{
"epoch": 20.41,
"learning_rate": 8.463589743589744e-06,
"loss": 0.0027,
"step": 3500
},
{
"epoch": 20.55,
"learning_rate": 8.450769230769231e-06,
"loss": 0.0032,
"step": 3525
},
{
"epoch": 20.7,
"learning_rate": 8.437948717948718e-06,
"loss": 0.0041,
"step": 3550
},
{
"epoch": 20.85,
"learning_rate": 8.425128205128205e-06,
"loss": 0.0037,
"step": 3575
},
{
"epoch": 20.99,
"learning_rate": 8.412307692307693e-06,
"loss": 0.0034,
"step": 3600
},
{
"epoch": 21.14,
"learning_rate": 8.39948717948718e-06,
"loss": 0.0037,
"step": 3625
},
{
"epoch": 21.28,
"learning_rate": 8.386666666666667e-06,
"loss": 0.003,
"step": 3650
},
{
"epoch": 21.43,
"learning_rate": 8.373846153846154e-06,
"loss": 0.0027,
"step": 3675
},
{
"epoch": 21.57,
"learning_rate": 8.361025641025642e-06,
"loss": 0.0025,
"step": 3700
},
{
"epoch": 21.72,
"learning_rate": 8.34820512820513e-06,
"loss": 0.0038,
"step": 3725
},
{
"epoch": 21.87,
"learning_rate": 8.335384615384616e-06,
"loss": 0.0044,
"step": 3750
},
{
"epoch": 22.01,
"learning_rate": 8.322564102564103e-06,
"loss": 0.0035,
"step": 3775
},
{
"epoch": 22.16,
"learning_rate": 8.309743589743591e-06,
"loss": 0.0033,
"step": 3800
},
{
"epoch": 22.3,
"learning_rate": 8.296923076923078e-06,
"loss": 0.0045,
"step": 3825
},
{
"epoch": 22.45,
"learning_rate": 8.284102564102565e-06,
"loss": 0.0035,
"step": 3850
},
{
"epoch": 22.59,
"learning_rate": 8.271282051282052e-06,
"loss": 0.0049,
"step": 3875
},
{
"epoch": 22.74,
"learning_rate": 8.25846153846154e-06,
"loss": 0.0037,
"step": 3900
},
{
"epoch": 22.89,
"learning_rate": 8.245641025641027e-06,
"loss": 0.0037,
"step": 3925
},
{
"epoch": 23.03,
"learning_rate": 8.232820512820512e-06,
"loss": 0.0031,
"step": 3950
},
{
"epoch": 23.18,
"learning_rate": 8.220000000000001e-06,
"loss": 0.0031,
"step": 3975
},
{
"epoch": 23.32,
"learning_rate": 8.207179487179488e-06,
"loss": 0.0029,
"step": 4000
},
{
"epoch": 23.32,
"eval_loss": 0.254440039396286,
"eval_runtime": 1315.9718,
"eval_samples_per_second": 4.861,
"eval_steps_per_second": 0.304,
"eval_wer": 6.5385812610244525,
"step": 4000
},
{
"epoch": 23.47,
"learning_rate": 8.194358974358975e-06,
"loss": 0.0031,
"step": 4025
},
{
"epoch": 23.62,
"learning_rate": 8.181538461538461e-06,
"loss": 0.0032,
"step": 4050
},
{
"epoch": 23.76,
"learning_rate": 8.16871794871795e-06,
"loss": 0.0036,
"step": 4075
},
{
"epoch": 23.91,
"learning_rate": 8.155897435897437e-06,
"loss": 0.0039,
"step": 4100
},
{
"epoch": 24.05,
"learning_rate": 8.143076923076924e-06,
"loss": 0.0036,
"step": 4125
},
{
"epoch": 24.2,
"learning_rate": 8.13025641025641e-06,
"loss": 0.0059,
"step": 4150
},
{
"epoch": 24.34,
"learning_rate": 8.117435897435897e-06,
"loss": 0.0048,
"step": 4175
},
{
"epoch": 24.49,
"learning_rate": 8.104615384615386e-06,
"loss": 0.0029,
"step": 4200
},
{
"epoch": 24.64,
"learning_rate": 8.091794871794873e-06,
"loss": 0.0032,
"step": 4225
},
{
"epoch": 24.78,
"learning_rate": 8.07897435897436e-06,
"loss": 0.003,
"step": 4250
},
{
"epoch": 24.93,
"learning_rate": 8.066153846153846e-06,
"loss": 0.0029,
"step": 4275
},
{
"epoch": 25.07,
"learning_rate": 8.053333333333335e-06,
"loss": 0.0028,
"step": 4300
},
{
"epoch": 25.22,
"learning_rate": 8.040512820512822e-06,
"loss": 0.0024,
"step": 4325
},
{
"epoch": 25.36,
"learning_rate": 8.027692307692308e-06,
"loss": 0.0028,
"step": 4350
},
{
"epoch": 25.51,
"learning_rate": 8.014871794871795e-06,
"loss": 0.0026,
"step": 4375
},
{
"epoch": 25.66,
"learning_rate": 8.002051282051284e-06,
"loss": 0.0034,
"step": 4400
},
{
"epoch": 25.8,
"learning_rate": 7.98923076923077e-06,
"loss": 0.0037,
"step": 4425
},
{
"epoch": 25.95,
"learning_rate": 7.976410256410257e-06,
"loss": 0.0024,
"step": 4450
},
{
"epoch": 26.09,
"learning_rate": 7.963589743589744e-06,
"loss": 0.0035,
"step": 4475
},
{
"epoch": 26.24,
"learning_rate": 7.950769230769233e-06,
"loss": 0.0028,
"step": 4500
},
{
"epoch": 26.38,
"learning_rate": 7.93794871794872e-06,
"loss": 0.0032,
"step": 4525
},
{
"epoch": 26.53,
"learning_rate": 7.925128205128205e-06,
"loss": 0.0032,
"step": 4550
},
{
"epoch": 26.68,
"learning_rate": 7.912307692307693e-06,
"loss": 0.0028,
"step": 4575
},
{
"epoch": 26.82,
"learning_rate": 7.89948717948718e-06,
"loss": 0.0032,
"step": 4600
},
{
"epoch": 26.97,
"learning_rate": 7.886666666666667e-06,
"loss": 0.0037,
"step": 4625
},
{
"epoch": 27.11,
"learning_rate": 7.873846153846154e-06,
"loss": 0.0025,
"step": 4650
},
{
"epoch": 27.26,
"learning_rate": 7.861025641025642e-06,
"loss": 0.0029,
"step": 4675
},
{
"epoch": 27.41,
"learning_rate": 7.848205128205129e-06,
"loss": 0.0026,
"step": 4700
},
{
"epoch": 27.55,
"learning_rate": 7.835384615384616e-06,
"loss": 0.0025,
"step": 4725
},
{
"epoch": 27.7,
"learning_rate": 7.822564102564103e-06,
"loss": 0.0017,
"step": 4750
},
{
"epoch": 27.84,
"learning_rate": 7.80974358974359e-06,
"loss": 0.0021,
"step": 4775
},
{
"epoch": 27.99,
"learning_rate": 7.796923076923078e-06,
"loss": 0.0023,
"step": 4800
},
{
"epoch": 28.13,
"learning_rate": 7.784102564102565e-06,
"loss": 0.0024,
"step": 4825
},
{
"epoch": 28.28,
"learning_rate": 7.771282051282052e-06,
"loss": 0.0027,
"step": 4850
},
{
"epoch": 28.43,
"learning_rate": 7.758461538461538e-06,
"loss": 0.0026,
"step": 4875
},
{
"epoch": 28.57,
"learning_rate": 7.745641025641027e-06,
"loss": 0.0023,
"step": 4900
},
{
"epoch": 28.72,
"learning_rate": 7.732820512820514e-06,
"loss": 0.0029,
"step": 4925
},
{
"epoch": 28.86,
"learning_rate": 7.72e-06,
"loss": 0.0029,
"step": 4950
},
{
"epoch": 29.01,
"learning_rate": 7.707179487179487e-06,
"loss": 0.003,
"step": 4975
},
{
"epoch": 29.15,
"learning_rate": 7.694358974358976e-06,
"loss": 0.0036,
"step": 5000
},
{
"epoch": 29.15,
"eval_loss": 0.25522172451019287,
"eval_runtime": 1316.2871,
"eval_samples_per_second": 4.86,
"eval_steps_per_second": 0.304,
"eval_wer": 6.697679244630443,
"step": 5000
},
{
"epoch": 29.3,
"learning_rate": 7.681538461538463e-06,
"loss": 0.0026,
"step": 5025
},
{
"epoch": 29.45,
"learning_rate": 7.66871794871795e-06,
"loss": 0.0025,
"step": 5050
},
{
"epoch": 29.59,
"learning_rate": 7.655897435897436e-06,
"loss": 0.0028,
"step": 5075
},
{
"epoch": 29.74,
"learning_rate": 7.643076923076925e-06,
"loss": 0.0025,
"step": 5100
},
{
"epoch": 29.88,
"learning_rate": 7.630256410256412e-06,
"loss": 0.002,
"step": 5125
},
{
"epoch": 30.03,
"learning_rate": 7.617435897435898e-06,
"loss": 0.002,
"step": 5150
},
{
"epoch": 30.17,
"learning_rate": 7.604615384615385e-06,
"loss": 0.0022,
"step": 5175
},
{
"epoch": 30.32,
"learning_rate": 7.591794871794872e-06,
"loss": 0.0025,
"step": 5200
},
{
"epoch": 30.47,
"learning_rate": 7.578974358974359e-06,
"loss": 0.0023,
"step": 5225
},
{
"epoch": 30.61,
"learning_rate": 7.566153846153847e-06,
"loss": 0.0021,
"step": 5250
},
{
"epoch": 30.76,
"learning_rate": 7.553333333333334e-06,
"loss": 0.0025,
"step": 5275
},
{
"epoch": 30.9,
"learning_rate": 7.540512820512821e-06,
"loss": 0.0026,
"step": 5300
},
{
"epoch": 31.05,
"learning_rate": 7.527692307692308e-06,
"loss": 0.0028,
"step": 5325
},
{
"epoch": 31.2,
"learning_rate": 7.514871794871795e-06,
"loss": 0.0018,
"step": 5350
},
{
"epoch": 31.34,
"learning_rate": 7.5020512820512826e-06,
"loss": 0.0023,
"step": 5375
},
{
"epoch": 31.49,
"learning_rate": 7.489230769230769e-06,
"loss": 0.0028,
"step": 5400
},
{
"epoch": 31.63,
"learning_rate": 7.476410256410257e-06,
"loss": 0.0019,
"step": 5425
},
{
"epoch": 31.78,
"learning_rate": 7.463589743589744e-06,
"loss": 0.0026,
"step": 5450
},
{
"epoch": 31.92,
"learning_rate": 7.4507692307692316e-06,
"loss": 0.0022,
"step": 5475
},
{
"epoch": 32.07,
"learning_rate": 7.437948717948718e-06,
"loss": 0.0025,
"step": 5500
},
{
"epoch": 32.22,
"learning_rate": 7.425128205128206e-06,
"loss": 0.0017,
"step": 5525
},
{
"epoch": 32.36,
"learning_rate": 7.412307692307693e-06,
"loss": 0.0011,
"step": 5550
},
{
"epoch": 32.51,
"learning_rate": 7.3994871794871806e-06,
"loss": 0.0021,
"step": 5575
},
{
"epoch": 32.65,
"learning_rate": 7.386666666666667e-06,
"loss": 0.0033,
"step": 5600
},
{
"epoch": 32.8,
"learning_rate": 7.373846153846155e-06,
"loss": 0.0017,
"step": 5625
},
{
"epoch": 32.94,
"learning_rate": 7.361025641025642e-06,
"loss": 0.0024,
"step": 5650
},
{
"epoch": 33.09,
"learning_rate": 7.3482051282051295e-06,
"loss": 0.0019,
"step": 5675
},
{
"epoch": 33.24,
"learning_rate": 7.335384615384616e-06,
"loss": 0.0021,
"step": 5700
},
{
"epoch": 33.38,
"learning_rate": 7.322564102564104e-06,
"loss": 0.0018,
"step": 5725
},
{
"epoch": 33.53,
"learning_rate": 7.309743589743591e-06,
"loss": 0.0017,
"step": 5750
},
{
"epoch": 33.67,
"learning_rate": 7.296923076923077e-06,
"loss": 0.0028,
"step": 5775
},
{
"epoch": 33.82,
"learning_rate": 7.2841025641025645e-06,
"loss": 0.002,
"step": 5800
},
{
"epoch": 33.97,
"learning_rate": 7.271282051282051e-06,
"loss": 0.0021,
"step": 5825
},
{
"epoch": 34.11,
"learning_rate": 7.258461538461539e-06,
"loss": 0.0028,
"step": 5850
},
{
"epoch": 34.26,
"learning_rate": 7.245641025641026e-06,
"loss": 0.0027,
"step": 5875
},
{
"epoch": 34.4,
"learning_rate": 7.2328205128205135e-06,
"loss": 0.002,
"step": 5900
},
{
"epoch": 34.55,
"learning_rate": 7.22e-06,
"loss": 0.002,
"step": 5925
},
{
"epoch": 34.69,
"learning_rate": 7.207179487179487e-06,
"loss": 0.0038,
"step": 5950
},
{
"epoch": 34.84,
"learning_rate": 7.194358974358975e-06,
"loss": 0.0024,
"step": 5975
},
{
"epoch": 34.99,
"learning_rate": 7.181538461538462e-06,
"loss": 0.0026,
"step": 6000
},
{
"epoch": 34.99,
"eval_loss": 0.27366775274276733,
"eval_runtime": 1315.5998,
"eval_samples_per_second": 4.862,
"eval_steps_per_second": 0.304,
"eval_wer": 6.856777228236434,
"step": 6000
},
{
"epoch": 35.13,
"learning_rate": 7.168717948717949e-06,
"loss": 0.0029,
"step": 6025
},
{
"epoch": 35.28,
"learning_rate": 7.155897435897436e-06,
"loss": 0.002,
"step": 6050
},
{
"epoch": 35.42,
"learning_rate": 7.143076923076924e-06,
"loss": 0.0018,
"step": 6075
},
{
"epoch": 35.57,
"learning_rate": 7.130256410256411e-06,
"loss": 0.0027,
"step": 6100
},
{
"epoch": 35.71,
"learning_rate": 7.117435897435898e-06,
"loss": 0.0028,
"step": 6125
},
{
"epoch": 35.86,
"learning_rate": 7.104615384615385e-06,
"loss": 0.0026,
"step": 6150
},
{
"epoch": 36.01,
"learning_rate": 7.091794871794873e-06,
"loss": 0.0039,
"step": 6175
},
{
"epoch": 36.15,
"learning_rate": 7.07897435897436e-06,
"loss": 0.0018,
"step": 6200
},
{
"epoch": 36.3,
"learning_rate": 7.066153846153847e-06,
"loss": 0.0017,
"step": 6225
},
{
"epoch": 36.44,
"learning_rate": 7.053333333333334e-06,
"loss": 0.0015,
"step": 6250
},
{
"epoch": 36.59,
"learning_rate": 7.040512820512822e-06,
"loss": 0.003,
"step": 6275
},
{
"epoch": 36.73,
"learning_rate": 7.027692307692309e-06,
"loss": 0.0018,
"step": 6300
},
{
"epoch": 36.88,
"learning_rate": 7.014871794871796e-06,
"loss": 0.0024,
"step": 6325
},
{
"epoch": 37.03,
"learning_rate": 7.002051282051283e-06,
"loss": 0.0036,
"step": 6350
},
{
"epoch": 37.17,
"learning_rate": 6.989230769230769e-06,
"loss": 0.0021,
"step": 6375
},
{
"epoch": 37.32,
"learning_rate": 6.976410256410257e-06,
"loss": 0.0019,
"step": 6400
},
{
"epoch": 37.46,
"learning_rate": 6.963589743589744e-06,
"loss": 0.0023,
"step": 6425
},
{
"epoch": 37.61,
"learning_rate": 6.950769230769231e-06,
"loss": 0.0025,
"step": 6450
},
{
"epoch": 37.76,
"learning_rate": 6.937948717948718e-06,
"loss": 0.0012,
"step": 6475
},
{
"epoch": 37.9,
"learning_rate": 6.925128205128206e-06,
"loss": 0.0019,
"step": 6500
},
{
"epoch": 38.05,
"learning_rate": 6.912307692307693e-06,
"loss": 0.0023,
"step": 6525
},
{
"epoch": 38.19,
"learning_rate": 6.899487179487179e-06,
"loss": 0.0021,
"step": 6550
},
{
"epoch": 38.34,
"learning_rate": 6.886666666666667e-06,
"loss": 0.0015,
"step": 6575
},
{
"epoch": 38.48,
"learning_rate": 6.873846153846154e-06,
"loss": 0.0011,
"step": 6600
},
{
"epoch": 38.63,
"learning_rate": 6.861025641025642e-06,
"loss": 0.0011,
"step": 6625
},
{
"epoch": 38.78,
"learning_rate": 6.848205128205128e-06,
"loss": 0.0016,
"step": 6650
},
{
"epoch": 38.92,
"learning_rate": 6.835384615384616e-06,
"loss": 0.0023,
"step": 6675
},
{
"epoch": 39.07,
"learning_rate": 6.822564102564103e-06,
"loss": 0.001,
"step": 6700
},
{
"epoch": 39.21,
"learning_rate": 6.8097435897435906e-06,
"loss": 0.0009,
"step": 6725
},
{
"epoch": 39.36,
"learning_rate": 6.796923076923077e-06,
"loss": 0.0017,
"step": 6750
},
{
"epoch": 39.5,
"learning_rate": 6.784102564102565e-06,
"loss": 0.0011,
"step": 6775
},
{
"epoch": 39.65,
"learning_rate": 6.771282051282052e-06,
"loss": 0.0015,
"step": 6800
},
{
"epoch": 39.8,
"learning_rate": 6.7584615384615396e-06,
"loss": 0.0014,
"step": 6825
},
{
"epoch": 39.94,
"learning_rate": 6.745641025641026e-06,
"loss": 0.0016,
"step": 6850
},
{
"epoch": 40.09,
"learning_rate": 6.732820512820514e-06,
"loss": 0.0012,
"step": 6875
},
{
"epoch": 40.23,
"learning_rate": 6.720000000000001e-06,
"loss": 0.0019,
"step": 6900
},
{
"epoch": 40.38,
"learning_rate": 6.7071794871794886e-06,
"loss": 0.0017,
"step": 6925
},
{
"epoch": 40.52,
"learning_rate": 6.694358974358975e-06,
"loss": 0.0018,
"step": 6950
},
{
"epoch": 40.67,
"learning_rate": 6.681538461538463e-06,
"loss": 0.0011,
"step": 6975
},
{
"epoch": 40.82,
"learning_rate": 6.668717948717949e-06,
"loss": 0.0009,
"step": 7000
},
{
"epoch": 40.82,
"eval_loss": 0.27335312962532043,
"eval_runtime": 1315.3258,
"eval_samples_per_second": 4.863,
"eval_steps_per_second": 0.304,
"eval_wer": 6.631964860097534,
"step": 7000
},
{
"epoch": 40.96,
"learning_rate": 6.655897435897436e-06,
"loss": 0.0018,
"step": 7025
},
{
"epoch": 41.11,
"learning_rate": 6.6430769230769235e-06,
"loss": 0.0013,
"step": 7050
},
{
"epoch": 41.25,
"learning_rate": 6.63025641025641e-06,
"loss": 0.0013,
"step": 7075
},
{
"epoch": 41.4,
"learning_rate": 6.617435897435898e-06,
"loss": 0.0019,
"step": 7100
},
{
"epoch": 41.55,
"learning_rate": 6.604615384615385e-06,
"loss": 0.0016,
"step": 7125
},
{
"epoch": 41.69,
"learning_rate": 6.5917948717948725e-06,
"loss": 0.0019,
"step": 7150
},
{
"epoch": 41.84,
"learning_rate": 6.578974358974359e-06,
"loss": 0.0022,
"step": 7175
},
{
"epoch": 41.98,
"learning_rate": 6.566153846153846e-06,
"loss": 0.0024,
"step": 7200
},
{
"epoch": 42.13,
"learning_rate": 6.553333333333334e-06,
"loss": 0.0015,
"step": 7225
},
{
"epoch": 42.27,
"learning_rate": 6.540512820512821e-06,
"loss": 0.0014,
"step": 7250
},
{
"epoch": 42.42,
"learning_rate": 6.527692307692308e-06,
"loss": 0.0017,
"step": 7275
},
{
"epoch": 42.57,
"learning_rate": 6.514871794871795e-06,
"loss": 0.003,
"step": 7300
},
{
"epoch": 42.71,
"learning_rate": 6.502051282051283e-06,
"loss": 0.0029,
"step": 7325
},
{
"epoch": 42.86,
"learning_rate": 6.48923076923077e-06,
"loss": 0.003,
"step": 7350
},
{
"epoch": 43.0,
"learning_rate": 6.476410256410257e-06,
"loss": 0.0031,
"step": 7375
},
{
"epoch": 43.15,
"learning_rate": 6.463589743589744e-06,
"loss": 0.0025,
"step": 7400
},
{
"epoch": 43.29,
"learning_rate": 6.450769230769232e-06,
"loss": 0.0018,
"step": 7425
},
{
"epoch": 43.44,
"learning_rate": 6.437948717948719e-06,
"loss": 0.0017,
"step": 7450
},
{
"epoch": 43.59,
"learning_rate": 6.425128205128206e-06,
"loss": 0.0022,
"step": 7475
},
{
"epoch": 43.73,
"learning_rate": 6.412307692307693e-06,
"loss": 0.0015,
"step": 7500
},
{
"epoch": 43.88,
"learning_rate": 6.399487179487181e-06,
"loss": 0.0023,
"step": 7525
},
{
"epoch": 44.02,
"learning_rate": 6.386666666666668e-06,
"loss": 0.0019,
"step": 7550
},
{
"epoch": 44.17,
"learning_rate": 6.373846153846155e-06,
"loss": 0.0015,
"step": 7575
},
{
"epoch": 44.31,
"learning_rate": 6.361025641025641e-06,
"loss": 0.0009,
"step": 7600
},
{
"epoch": 44.46,
"learning_rate": 6.348205128205128e-06,
"loss": 0.0011,
"step": 7625
},
{
"epoch": 44.61,
"learning_rate": 6.335384615384616e-06,
"loss": 0.0012,
"step": 7650
},
{
"epoch": 44.75,
"learning_rate": 6.322564102564103e-06,
"loss": 0.0018,
"step": 7675
},
{
"epoch": 44.9,
"learning_rate": 6.30974358974359e-06,
"loss": 0.0015,
"step": 7700
},
{
"epoch": 45.04,
"learning_rate": 6.296923076923077e-06,
"loss": 0.0008,
"step": 7725
},
{
"epoch": 45.19,
"learning_rate": 6.284102564102565e-06,
"loss": 0.0008,
"step": 7750
},
{
"epoch": 45.34,
"learning_rate": 6.271282051282052e-06,
"loss": 0.0012,
"step": 7775
},
{
"epoch": 45.48,
"learning_rate": 6.258461538461538e-06,
"loss": 0.001,
"step": 7800
},
{
"epoch": 45.63,
"learning_rate": 6.245641025641026e-06,
"loss": 0.0009,
"step": 7825
},
{
"epoch": 45.77,
"learning_rate": 6.232820512820513e-06,
"loss": 0.0012,
"step": 7850
},
{
"epoch": 45.92,
"learning_rate": 6.220000000000001e-06,
"loss": 0.0009,
"step": 7875
},
{
"epoch": 46.06,
"learning_rate": 6.207179487179487e-06,
"loss": 0.0013,
"step": 7900
},
{
"epoch": 46.21,
"learning_rate": 6.194358974358975e-06,
"loss": 0.0017,
"step": 7925
},
{
"epoch": 46.36,
"learning_rate": 6.181538461538462e-06,
"loss": 0.0017,
"step": 7950
},
{
"epoch": 46.5,
"learning_rate": 6.16871794871795e-06,
"loss": 0.0014,
"step": 7975
},
{
"epoch": 46.65,
"learning_rate": 6.155897435897436e-06,
"loss": 0.0009,
"step": 8000
},
{
"epoch": 46.65,
"eval_loss": 0.2768790125846863,
"eval_runtime": 1314.5097,
"eval_samples_per_second": 4.866,
"eval_steps_per_second": 0.304,
"eval_wer": 6.818732058243697,
"step": 8000
},
{
"epoch": 46.79,
"learning_rate": 6.143076923076924e-06,
"loss": 0.0012,
"step": 8025
},
{
"epoch": 46.94,
"learning_rate": 6.130256410256411e-06,
"loss": 0.0015,
"step": 8050
},
{
"epoch": 47.08,
"learning_rate": 6.1174358974358986e-06,
"loss": 0.0014,
"step": 8075
},
{
"epoch": 47.23,
"learning_rate": 6.104615384615385e-06,
"loss": 0.001,
"step": 8100
},
{
"epoch": 47.38,
"learning_rate": 6.091794871794873e-06,
"loss": 0.0006,
"step": 8125
},
{
"epoch": 47.52,
"learning_rate": 6.07897435897436e-06,
"loss": 0.0005,
"step": 8150
},
{
"epoch": 47.67,
"learning_rate": 6.0661538461538476e-06,
"loss": 0.0009,
"step": 8175
},
{
"epoch": 47.81,
"learning_rate": 6.0533333333333335e-06,
"loss": 0.0009,
"step": 8200
},
{
"epoch": 47.96,
"learning_rate": 6.04051282051282e-06,
"loss": 0.0012,
"step": 8225
},
{
"epoch": 48.1,
"learning_rate": 6.027692307692308e-06,
"loss": 0.0016,
"step": 8250
},
{
"epoch": 48.25,
"learning_rate": 6.014871794871795e-06,
"loss": 0.0012,
"step": 8275
},
{
"epoch": 48.4,
"learning_rate": 6.0020512820512825e-06,
"loss": 0.0013,
"step": 8300
},
{
"epoch": 48.54,
"learning_rate": 5.989230769230769e-06,
"loss": 0.0012,
"step": 8325
},
{
"epoch": 48.69,
"learning_rate": 5.976410256410257e-06,
"loss": 0.0012,
"step": 8350
},
{
"epoch": 48.83,
"learning_rate": 5.963589743589744e-06,
"loss": 0.0013,
"step": 8375
},
{
"epoch": 48.98,
"learning_rate": 5.950769230769231e-06,
"loss": 0.0011,
"step": 8400
},
{
"epoch": 49.13,
"learning_rate": 5.937948717948718e-06,
"loss": 0.001,
"step": 8425
},
{
"epoch": 49.27,
"learning_rate": 5.925128205128205e-06,
"loss": 0.002,
"step": 8450
},
{
"epoch": 49.42,
"learning_rate": 5.912307692307693e-06,
"loss": 0.0013,
"step": 8475
},
{
"epoch": 49.56,
"learning_rate": 5.89948717948718e-06,
"loss": 0.001,
"step": 8500
},
{
"epoch": 49.71,
"learning_rate": 5.886666666666667e-06,
"loss": 0.0012,
"step": 8525
},
{
"epoch": 49.85,
"learning_rate": 5.873846153846154e-06,
"loss": 0.0012,
"step": 8550
},
{
"epoch": 50.0,
"learning_rate": 5.861025641025642e-06,
"loss": 0.0015,
"step": 8575
},
{
"epoch": 50.15,
"learning_rate": 5.848205128205129e-06,
"loss": 0.0008,
"step": 8600
},
{
"epoch": 50.29,
"learning_rate": 5.835384615384616e-06,
"loss": 0.0012,
"step": 8625
},
{
"epoch": 50.44,
"learning_rate": 5.822564102564103e-06,
"loss": 0.0013,
"step": 8650
},
{
"epoch": 50.58,
"learning_rate": 5.809743589743591e-06,
"loss": 0.0007,
"step": 8675
},
{
"epoch": 50.73,
"learning_rate": 5.796923076923078e-06,
"loss": 0.0007,
"step": 8700
},
{
"epoch": 50.87,
"learning_rate": 5.784102564102565e-06,
"loss": 0.0005,
"step": 8725
},
{
"epoch": 51.02,
"learning_rate": 5.771282051282052e-06,
"loss": 0.0007,
"step": 8750
},
{
"epoch": 51.17,
"learning_rate": 5.75846153846154e-06,
"loss": 0.0016,
"step": 8775
},
{
"epoch": 51.31,
"learning_rate": 5.745641025641027e-06,
"loss": 0.0013,
"step": 8800
},
{
"epoch": 51.46,
"learning_rate": 5.732820512820513e-06,
"loss": 0.0006,
"step": 8825
},
{
"epoch": 51.6,
"learning_rate": 5.72e-06,
"loss": 0.0008,
"step": 8850
},
{
"epoch": 51.75,
"learning_rate": 5.707179487179487e-06,
"loss": 0.0008,
"step": 8875
},
{
"epoch": 51.9,
"learning_rate": 5.694358974358975e-06,
"loss": 0.0008,
"step": 8900
},
{
"epoch": 52.04,
"learning_rate": 5.681538461538462e-06,
"loss": 0.0006,
"step": 8925
},
{
"epoch": 52.19,
"learning_rate": 5.668717948717949e-06,
"loss": 0.0008,
"step": 8950
},
{
"epoch": 52.33,
"learning_rate": 5.655897435897436e-06,
"loss": 0.0006,
"step": 8975
},
{
"epoch": 52.48,
"learning_rate": 5.643076923076923e-06,
"loss": 0.0006,
"step": 9000
},
{
"epoch": 52.48,
"eval_loss": 0.2831883132457733,
"eval_runtime": 1315.74,
"eval_samples_per_second": 4.862,
"eval_steps_per_second": 0.304,
"eval_wer": 6.616400926918688,
"step": 9000
},
{
"epoch": 52.62,
"learning_rate": 5.630256410256411e-06,
"loss": 0.0004,
"step": 9025
},
{
"epoch": 52.77,
"learning_rate": 5.6174358974358974e-06,
"loss": 0.0017,
"step": 9050
},
{
"epoch": 52.92,
"learning_rate": 5.604615384615385e-06,
"loss": 0.0007,
"step": 9075
},
{
"epoch": 53.06,
"learning_rate": 5.591794871794872e-06,
"loss": 0.0006,
"step": 9100
},
{
"epoch": 53.21,
"learning_rate": 5.57897435897436e-06,
"loss": 0.001,
"step": 9125
},
{
"epoch": 53.35,
"learning_rate": 5.566153846153846e-06,
"loss": 0.0011,
"step": 9150
},
{
"epoch": 53.5,
"learning_rate": 5.553333333333334e-06,
"loss": 0.0008,
"step": 9175
},
{
"epoch": 53.64,
"learning_rate": 5.540512820512821e-06,
"loss": 0.0006,
"step": 9200
},
{
"epoch": 53.79,
"learning_rate": 5.527692307692309e-06,
"loss": 0.0008,
"step": 9225
},
{
"epoch": 53.94,
"learning_rate": 5.514871794871795e-06,
"loss": 0.0005,
"step": 9250
},
{
"epoch": 54.08,
"learning_rate": 5.502051282051283e-06,
"loss": 0.0023,
"step": 9275
},
{
"epoch": 54.23,
"learning_rate": 5.48923076923077e-06,
"loss": 0.0006,
"step": 9300
},
{
"epoch": 54.37,
"learning_rate": 5.476410256410258e-06,
"loss": 0.0026,
"step": 9325
},
{
"epoch": 54.52,
"learning_rate": 5.463589743589744e-06,
"loss": 0.0012,
"step": 9350
},
{
"epoch": 54.66,
"learning_rate": 5.450769230769232e-06,
"loss": 0.0007,
"step": 9375
},
{
"epoch": 54.81,
"learning_rate": 5.437948717948719e-06,
"loss": 0.0006,
"step": 9400
},
{
"epoch": 54.96,
"learning_rate": 5.425128205128205e-06,
"loss": 0.0007,
"step": 9425
},
{
"epoch": 55.1,
"learning_rate": 5.4123076923076925e-06,
"loss": 0.0008,
"step": 9450
},
{
"epoch": 55.25,
"learning_rate": 5.399487179487179e-06,
"loss": 0.0013,
"step": 9475
},
{
"epoch": 55.39,
"learning_rate": 5.386666666666667e-06,
"loss": 0.0008,
"step": 9500
},
{
"epoch": 55.54,
"learning_rate": 5.373846153846154e-06,
"loss": 0.0006,
"step": 9525
},
{
"epoch": 55.69,
"learning_rate": 5.3610256410256415e-06,
"loss": 0.0011,
"step": 9550
},
{
"epoch": 55.83,
"learning_rate": 5.348205128205128e-06,
"loss": 0.0011,
"step": 9575
},
{
"epoch": 55.98,
"learning_rate": 5.335384615384615e-06,
"loss": 0.0007,
"step": 9600
},
{
"epoch": 56.12,
"learning_rate": 5.322564102564103e-06,
"loss": 0.0006,
"step": 9625
},
{
"epoch": 56.27,
"learning_rate": 5.30974358974359e-06,
"loss": 0.001,
"step": 9650
},
{
"epoch": 56.41,
"learning_rate": 5.296923076923077e-06,
"loss": 0.001,
"step": 9675
},
{
"epoch": 56.56,
"learning_rate": 5.284102564102564e-06,
"loss": 0.0011,
"step": 9700
},
{
"epoch": 56.71,
"learning_rate": 5.271282051282052e-06,
"loss": 0.0017,
"step": 9725
},
{
"epoch": 56.85,
"learning_rate": 5.258461538461539e-06,
"loss": 0.0008,
"step": 9750
},
{
"epoch": 57.0,
"learning_rate": 5.245641025641026e-06,
"loss": 0.0011,
"step": 9775
},
{
"epoch": 57.14,
"learning_rate": 5.232820512820513e-06,
"loss": 0.001,
"step": 9800
},
{
"epoch": 57.29,
"learning_rate": 5.220000000000001e-06,
"loss": 0.0008,
"step": 9825
},
{
"epoch": 57.43,
"learning_rate": 5.207179487179488e-06,
"loss": 0.0011,
"step": 9850
},
{
"epoch": 57.58,
"learning_rate": 5.194358974358975e-06,
"loss": 0.0012,
"step": 9875
},
{
"epoch": 57.73,
"learning_rate": 5.181538461538462e-06,
"loss": 0.0007,
"step": 9900
},
{
"epoch": 57.87,
"learning_rate": 5.16871794871795e-06,
"loss": 0.0017,
"step": 9925
},
{
"epoch": 58.02,
"learning_rate": 5.155897435897437e-06,
"loss": 0.002,
"step": 9950
},
{
"epoch": 58.16,
"learning_rate": 5.143076923076924e-06,
"loss": 0.0008,
"step": 9975
},
{
"epoch": 58.31,
"learning_rate": 5.130256410256411e-06,
"loss": 0.0013,
"step": 10000
},
{
"epoch": 58.31,
"eval_loss": 0.28828713297843933,
"eval_runtime": 1315.4767,
"eval_samples_per_second": 4.863,
"eval_steps_per_second": 0.304,
"eval_wer": 7.017604537751184,
"step": 10000
},
{
"epoch": 58.45,
"learning_rate": 5.117435897435897e-06,
"loss": 0.0013,
"step": 10025
},
{
"epoch": 58.6,
"learning_rate": 5.104615384615385e-06,
"loss": 0.0011,
"step": 10050
},
{
"epoch": 58.75,
"learning_rate": 5.091794871794872e-06,
"loss": 0.0007,
"step": 10075
},
{
"epoch": 58.89,
"learning_rate": 5.078974358974359e-06,
"loss": 0.0009,
"step": 10100
},
{
"epoch": 59.04,
"learning_rate": 5.066153846153846e-06,
"loss": 0.0011,
"step": 10125
},
{
"epoch": 59.18,
"learning_rate": 5.053333333333334e-06,
"loss": 0.001,
"step": 10150
},
{
"epoch": 59.33,
"learning_rate": 5.040512820512821e-06,
"loss": 0.0008,
"step": 10175
},
{
"epoch": 59.48,
"learning_rate": 5.027692307692308e-06,
"loss": 0.0009,
"step": 10200
},
{
"epoch": 59.62,
"learning_rate": 5.014871794871795e-06,
"loss": 0.0006,
"step": 10225
},
{
"epoch": 59.77,
"learning_rate": 5.002051282051282e-06,
"loss": 0.0011,
"step": 10250
},
{
"epoch": 59.91,
"learning_rate": 4.98923076923077e-06,
"loss": 0.0012,
"step": 10275
},
{
"epoch": 60.06,
"learning_rate": 4.9764102564102564e-06,
"loss": 0.0015,
"step": 10300
},
{
"epoch": 60.2,
"learning_rate": 4.963589743589744e-06,
"loss": 0.0006,
"step": 10325
},
{
"epoch": 60.35,
"learning_rate": 4.950769230769231e-06,
"loss": 0.0014,
"step": 10350
},
{
"epoch": 60.5,
"learning_rate": 4.937948717948719e-06,
"loss": 0.0007,
"step": 10375
},
{
"epoch": 60.64,
"learning_rate": 4.9251282051282054e-06,
"loss": 0.001,
"step": 10400
},
{
"epoch": 60.79,
"learning_rate": 4.912307692307693e-06,
"loss": 0.0008,
"step": 10425
},
{
"epoch": 60.93,
"learning_rate": 4.89948717948718e-06,
"loss": 0.0007,
"step": 10450
},
{
"epoch": 61.08,
"learning_rate": 4.886666666666668e-06,
"loss": 0.0017,
"step": 10475
},
{
"epoch": 61.22,
"learning_rate": 4.873846153846154e-06,
"loss": 0.0012,
"step": 10500
},
{
"epoch": 61.37,
"learning_rate": 4.861025641025641e-06,
"loss": 0.0013,
"step": 10525
},
{
"epoch": 61.52,
"learning_rate": 4.848205128205128e-06,
"loss": 0.0006,
"step": 10550
},
{
"epoch": 61.66,
"learning_rate": 4.835384615384616e-06,
"loss": 0.0011,
"step": 10575
},
{
"epoch": 61.81,
"learning_rate": 4.8225641025641026e-06,
"loss": 0.0012,
"step": 10600
},
{
"epoch": 61.95,
"learning_rate": 4.80974358974359e-06,
"loss": 0.0008,
"step": 10625
},
{
"epoch": 62.1,
"learning_rate": 4.796923076923077e-06,
"loss": 0.0009,
"step": 10650
},
{
"epoch": 62.24,
"learning_rate": 4.784102564102565e-06,
"loss": 0.0008,
"step": 10675
},
{
"epoch": 62.39,
"learning_rate": 4.7712820512820516e-06,
"loss": 0.0013,
"step": 10700
},
{
"epoch": 62.54,
"learning_rate": 4.758461538461539e-06,
"loss": 0.0015,
"step": 10725
},
{
"epoch": 62.68,
"learning_rate": 4.745641025641026e-06,
"loss": 0.0006,
"step": 10750
},
{
"epoch": 62.83,
"learning_rate": 4.732820512820514e-06,
"loss": 0.0013,
"step": 10775
},
{
"epoch": 62.97,
"learning_rate": 4.7200000000000005e-06,
"loss": 0.0009,
"step": 10800
},
{
"epoch": 63.12,
"learning_rate": 4.707179487179487e-06,
"loss": 0.0006,
"step": 10825
},
{
"epoch": 63.27,
"learning_rate": 4.694358974358974e-06,
"loss": 0.0005,
"step": 10850
},
{
"epoch": 63.41,
"learning_rate": 4.681538461538462e-06,
"loss": 0.0009,
"step": 10875
},
{
"epoch": 63.56,
"learning_rate": 4.668717948717949e-06,
"loss": 0.0003,
"step": 10900
},
{
"epoch": 63.7,
"learning_rate": 4.655897435897436e-06,
"loss": 0.0005,
"step": 10925
},
{
"epoch": 63.85,
"learning_rate": 4.643076923076923e-06,
"loss": 0.001,
"step": 10950
},
{
"epoch": 63.99,
"learning_rate": 4.630256410256411e-06,
"loss": 0.0011,
"step": 10975
},
{
"epoch": 64.14,
"learning_rate": 4.617435897435898e-06,
"loss": 0.0005,
"step": 11000
},
{
"epoch": 64.14,
"eval_loss": 0.29717591404914856,
"eval_runtime": 1315.8176,
"eval_samples_per_second": 4.862,
"eval_steps_per_second": 0.304,
"eval_wer": 6.898281050046692,
"step": 11000
},
{
"epoch": 64.29,
"learning_rate": 4.604615384615385e-06,
"loss": 0.0008,
"step": 11025
},
{
"epoch": 64.43,
"learning_rate": 4.591794871794872e-06,
"loss": 0.0004,
"step": 11050
},
{
"epoch": 64.58,
"learning_rate": 4.57897435897436e-06,
"loss": 0.0007,
"step": 11075
},
{
"epoch": 64.72,
"learning_rate": 4.566153846153847e-06,
"loss": 0.0005,
"step": 11100
},
{
"epoch": 64.87,
"learning_rate": 4.5533333333333335e-06,
"loss": 0.0003,
"step": 11125
},
{
"epoch": 65.01,
"learning_rate": 4.54051282051282e-06,
"loss": 0.0005,
"step": 11150
},
{
"epoch": 65.16,
"learning_rate": 4.527692307692308e-06,
"loss": 0.0002,
"step": 11175
},
{
"epoch": 65.31,
"learning_rate": 4.514871794871795e-06,
"loss": 0.0007,
"step": 11200
},
{
"epoch": 65.45,
"learning_rate": 4.5020512820512825e-06,
"loss": 0.0005,
"step": 11225
},
{
"epoch": 65.6,
"learning_rate": 4.489230769230769e-06,
"loss": 0.0004,
"step": 11250
},
{
"epoch": 65.74,
"learning_rate": 4.476410256410257e-06,
"loss": 0.0008,
"step": 11275
},
{
"epoch": 65.89,
"learning_rate": 4.463589743589744e-06,
"loss": 0.0007,
"step": 11300
},
{
"epoch": 66.03,
"learning_rate": 4.4507692307692315e-06,
"loss": 0.0007,
"step": 11325
},
{
"epoch": 66.18,
"learning_rate": 4.437948717948718e-06,
"loss": 0.0005,
"step": 11350
},
{
"epoch": 66.33,
"learning_rate": 4.425128205128206e-06,
"loss": 0.0009,
"step": 11375
},
{
"epoch": 66.47,
"learning_rate": 4.412307692307693e-06,
"loss": 0.0004,
"step": 11400
},
{
"epoch": 66.62,
"learning_rate": 4.39948717948718e-06,
"loss": 0.0004,
"step": 11425
},
{
"epoch": 66.76,
"learning_rate": 4.3866666666666665e-06,
"loss": 0.0005,
"step": 11450
},
{
"epoch": 66.91,
"learning_rate": 4.373846153846154e-06,
"loss": 0.0009,
"step": 11475
},
{
"epoch": 67.06,
"learning_rate": 4.361025641025641e-06,
"loss": 0.0004,
"step": 11500
},
{
"epoch": 67.2,
"learning_rate": 4.348205128205129e-06,
"loss": 0.0005,
"step": 11525
},
{
"epoch": 67.35,
"learning_rate": 4.3353846153846154e-06,
"loss": 0.0002,
"step": 11550
},
{
"epoch": 67.49,
"learning_rate": 4.322564102564103e-06,
"loss": 0.0004,
"step": 11575
},
{
"epoch": 67.64,
"learning_rate": 4.30974358974359e-06,
"loss": 0.0006,
"step": 11600
},
{
"epoch": 67.78,
"learning_rate": 4.296923076923078e-06,
"loss": 0.0002,
"step": 11625
},
{
"epoch": 67.93,
"learning_rate": 4.2841025641025644e-06,
"loss": 0.0004,
"step": 11650
},
{
"epoch": 68.08,
"learning_rate": 4.271282051282052e-06,
"loss": 0.0004,
"step": 11675
},
{
"epoch": 68.22,
"learning_rate": 4.258461538461539e-06,
"loss": 0.0004,
"step": 11700
},
{
"epoch": 68.37,
"learning_rate": 4.245641025641026e-06,
"loss": 0.0007,
"step": 11725
},
{
"epoch": 68.51,
"learning_rate": 4.232820512820513e-06,
"loss": 0.0002,
"step": 11750
},
{
"epoch": 68.66,
"learning_rate": 4.22e-06,
"loss": 0.0005,
"step": 11775
},
{
"epoch": 68.8,
"learning_rate": 4.207179487179487e-06,
"loss": 0.0002,
"step": 11800
},
{
"epoch": 68.95,
"learning_rate": 4.194358974358975e-06,
"loss": 0.0011,
"step": 11825
},
{
"epoch": 69.1,
"learning_rate": 4.1815384615384616e-06,
"loss": 0.0004,
"step": 11850
},
{
"epoch": 69.24,
"learning_rate": 4.168717948717949e-06,
"loss": 0.0011,
"step": 11875
},
{
"epoch": 69.39,
"learning_rate": 4.155897435897436e-06,
"loss": 0.0009,
"step": 11900
},
{
"epoch": 69.53,
"learning_rate": 4.143076923076924e-06,
"loss": 0.0004,
"step": 11925
},
{
"epoch": 69.68,
"learning_rate": 4.1302564102564106e-06,
"loss": 0.0003,
"step": 11950
},
{
"epoch": 69.83,
"learning_rate": 4.117435897435898e-06,
"loss": 0.0007,
"step": 11975
},
{
"epoch": 69.97,
"learning_rate": 4.104615384615385e-06,
"loss": 0.0006,
"step": 12000
},
{
"epoch": 69.97,
"eval_loss": 0.29644352197647095,
"eval_runtime": 1316.0527,
"eval_samples_per_second": 4.861,
"eval_steps_per_second": 0.304,
"eval_wer": 6.673468681907792,
"step": 12000
},
{
"epoch": 70.12,
"learning_rate": 4.091794871794872e-06,
"loss": 0.0003,
"step": 12025
},
{
"epoch": 70.26,
"learning_rate": 4.078974358974359e-06,
"loss": 0.0002,
"step": 12050
},
{
"epoch": 70.41,
"learning_rate": 4.066153846153846e-06,
"loss": 0.0003,
"step": 12075
},
{
"epoch": 70.55,
"learning_rate": 4.053333333333333e-06,
"loss": 0.0004,
"step": 12100
},
{
"epoch": 70.7,
"learning_rate": 4.040512820512821e-06,
"loss": 0.0003,
"step": 12125
},
{
"epoch": 70.85,
"learning_rate": 4.027692307692308e-06,
"loss": 0.0005,
"step": 12150
},
{
"epoch": 70.99,
"learning_rate": 4.014871794871795e-06,
"loss": 0.0005,
"step": 12175
},
{
"epoch": 71.14,
"learning_rate": 4.002051282051282e-06,
"loss": 0.0008,
"step": 12200
},
{
"epoch": 71.28,
"learning_rate": 3.98923076923077e-06,
"loss": 0.0003,
"step": 12225
},
{
"epoch": 71.43,
"learning_rate": 3.976410256410257e-06,
"loss": 0.0006,
"step": 12250
},
{
"epoch": 71.57,
"learning_rate": 3.963589743589744e-06,
"loss": 0.0003,
"step": 12275
},
{
"epoch": 71.72,
"learning_rate": 3.950769230769231e-06,
"loss": 0.0002,
"step": 12300
},
{
"epoch": 71.87,
"learning_rate": 3.937948717948718e-06,
"loss": 0.0002,
"step": 12325
},
{
"epoch": 72.01,
"learning_rate": 3.925128205128205e-06,
"loss": 0.0003,
"step": 12350
},
{
"epoch": 72.16,
"learning_rate": 3.9123076923076925e-06,
"loss": 0.0003,
"step": 12375
},
{
"epoch": 72.3,
"learning_rate": 3.899487179487179e-06,
"loss": 0.0006,
"step": 12400
},
{
"epoch": 72.45,
"learning_rate": 3.886666666666667e-06,
"loss": 0.0006,
"step": 12425
},
{
"epoch": 72.59,
"learning_rate": 3.873846153846154e-06,
"loss": 0.0007,
"step": 12450
},
{
"epoch": 72.74,
"learning_rate": 3.8610256410256415e-06,
"loss": 0.0003,
"step": 12475
},
{
"epoch": 72.89,
"learning_rate": 3.848205128205128e-06,
"loss": 0.0004,
"step": 12500
},
{
"epoch": 73.03,
"learning_rate": 3.835384615384616e-06,
"loss": 0.0005,
"step": 12525
},
{
"epoch": 73.18,
"learning_rate": 3.822564102564103e-06,
"loss": 0.0002,
"step": 12550
},
{
"epoch": 73.32,
"learning_rate": 3.80974358974359e-06,
"loss": 0.0003,
"step": 12575
},
{
"epoch": 73.47,
"learning_rate": 3.7969230769230773e-06,
"loss": 0.0004,
"step": 12600
},
{
"epoch": 73.62,
"learning_rate": 3.784102564102564e-06,
"loss": 0.0005,
"step": 12625
},
{
"epoch": 73.76,
"learning_rate": 3.7712820512820514e-06,
"loss": 0.0007,
"step": 12650
},
{
"epoch": 73.91,
"learning_rate": 3.7584615384615386e-06,
"loss": 0.0002,
"step": 12675
},
{
"epoch": 74.05,
"learning_rate": 3.745641025641026e-06,
"loss": 0.0006,
"step": 12700
},
{
"epoch": 74.2,
"learning_rate": 3.732820512820513e-06,
"loss": 0.0004,
"step": 12725
},
{
"epoch": 74.34,
"learning_rate": 3.7200000000000004e-06,
"loss": 0.0003,
"step": 12750
},
{
"epoch": 74.49,
"learning_rate": 3.7071794871794876e-06,
"loss": 0.0006,
"step": 12775
},
{
"epoch": 74.64,
"learning_rate": 3.694358974358975e-06,
"loss": 0.0003,
"step": 12800
},
{
"epoch": 74.78,
"learning_rate": 3.681538461538462e-06,
"loss": 0.0006,
"step": 12825
},
{
"epoch": 74.93,
"learning_rate": 3.6687179487179494e-06,
"loss": 0.0006,
"step": 12850
},
{
"epoch": 75.07,
"learning_rate": 3.655897435897436e-06,
"loss": 0.0003,
"step": 12875
},
{
"epoch": 75.22,
"learning_rate": 3.6430769230769234e-06,
"loss": 0.0003,
"step": 12900
},
{
"epoch": 75.36,
"learning_rate": 3.6302564102564103e-06,
"loss": 0.0005,
"step": 12925
},
{
"epoch": 75.51,
"learning_rate": 3.6174358974358975e-06,
"loss": 0.0005,
"step": 12950
},
{
"epoch": 75.66,
"learning_rate": 3.6046153846153848e-06,
"loss": 0.0003,
"step": 12975
},
{
"epoch": 75.8,
"learning_rate": 3.591794871794872e-06,
"loss": 0.0003,
"step": 13000
},
{
"epoch": 75.8,
"eval_loss": 0.30420011281967163,
"eval_runtime": 1315.9629,
"eval_samples_per_second": 4.861,
"eval_steps_per_second": 0.304,
"eval_wer": 6.7391830664407015,
"step": 13000
},
{
"epoch": 75.95,
"learning_rate": 3.5789743589743593e-06,
"loss": 0.0005,
"step": 13025
},
{
"epoch": 76.09,
"learning_rate": 3.5661538461538465e-06,
"loss": 0.0005,
"step": 13050
},
{
"epoch": 76.24,
"learning_rate": 3.5533333333333338e-06,
"loss": 0.0002,
"step": 13075
},
{
"epoch": 76.38,
"learning_rate": 3.540512820512821e-06,
"loss": 0.0006,
"step": 13100
},
{
"epoch": 76.53,
"learning_rate": 3.5276923076923083e-06,
"loss": 0.0004,
"step": 13125
},
{
"epoch": 76.68,
"learning_rate": 3.5148717948717955e-06,
"loss": 0.0002,
"step": 13150
},
{
"epoch": 76.82,
"learning_rate": 3.5020512820512823e-06,
"loss": 0.0006,
"step": 13175
},
{
"epoch": 76.97,
"learning_rate": 3.4892307692307696e-06,
"loss": 0.0005,
"step": 13200
},
{
"epoch": 77.11,
"learning_rate": 3.4764102564102564e-06,
"loss": 0.0002,
"step": 13225
},
{
"epoch": 77.26,
"learning_rate": 3.4635897435897436e-06,
"loss": 0.0002,
"step": 13250
},
{
"epoch": 77.41,
"learning_rate": 3.450769230769231e-06,
"loss": 0.0002,
"step": 13275
},
{
"epoch": 77.55,
"learning_rate": 3.4384615384615387e-06,
"loss": 0.0004,
"step": 13300
},
{
"epoch": 77.7,
"learning_rate": 3.425641025641026e-06,
"loss": 0.0003,
"step": 13325
},
{
"epoch": 77.84,
"learning_rate": 3.4128205128205132e-06,
"loss": 0.0003,
"step": 13350
},
{
"epoch": 77.99,
"learning_rate": 3.4000000000000005e-06,
"loss": 0.0002,
"step": 13375
},
{
"epoch": 78.13,
"learning_rate": 3.3871794871794873e-06,
"loss": 0.0004,
"step": 13400
},
{
"epoch": 78.28,
"learning_rate": 3.3743589743589746e-06,
"loss": 0.0006,
"step": 13425
},
{
"epoch": 78.43,
"learning_rate": 3.361538461538462e-06,
"loss": 0.0003,
"step": 13450
},
{
"epoch": 78.57,
"learning_rate": 3.348717948717949e-06,
"loss": 0.0002,
"step": 13475
},
{
"epoch": 78.72,
"learning_rate": 3.3358974358974363e-06,
"loss": 0.0002,
"step": 13500
},
{
"epoch": 78.86,
"learning_rate": 3.323076923076923e-06,
"loss": 0.0002,
"step": 13525
},
{
"epoch": 79.01,
"learning_rate": 3.3102564102564104e-06,
"loss": 0.0002,
"step": 13550
},
{
"epoch": 79.15,
"learning_rate": 3.2974358974358976e-06,
"loss": 0.0003,
"step": 13575
},
{
"epoch": 79.3,
"learning_rate": 3.284615384615385e-06,
"loss": 0.0003,
"step": 13600
},
{
"epoch": 79.45,
"learning_rate": 3.271794871794872e-06,
"loss": 0.0002,
"step": 13625
},
{
"epoch": 79.59,
"learning_rate": 3.2589743589743594e-06,
"loss": 0.0009,
"step": 13650
},
{
"epoch": 79.74,
"learning_rate": 3.2461538461538466e-06,
"loss": 0.0005,
"step": 13675
},
{
"epoch": 79.88,
"learning_rate": 3.2333333333333334e-06,
"loss": 0.0006,
"step": 13700
},
{
"epoch": 80.03,
"learning_rate": 3.2205128205128207e-06,
"loss": 0.0002,
"step": 13725
},
{
"epoch": 80.17,
"learning_rate": 3.207692307692308e-06,
"loss": 0.0002,
"step": 13750
},
{
"epoch": 80.32,
"learning_rate": 3.194871794871795e-06,
"loss": 0.0005,
"step": 13775
},
{
"epoch": 80.47,
"learning_rate": 3.1820512820512824e-06,
"loss": 0.0005,
"step": 13800
},
{
"epoch": 80.61,
"learning_rate": 3.1692307692307693e-06,
"loss": 0.0006,
"step": 13825
},
{
"epoch": 80.76,
"learning_rate": 3.1564102564102565e-06,
"loss": 0.0002,
"step": 13850
},
{
"epoch": 80.9,
"learning_rate": 3.1435897435897437e-06,
"loss": 0.0002,
"step": 13875
},
{
"epoch": 81.05,
"learning_rate": 3.130769230769231e-06,
"loss": 0.0004,
"step": 13900
},
{
"epoch": 81.2,
"learning_rate": 3.1179487179487182e-06,
"loss": 0.0001,
"step": 13925
},
{
"epoch": 81.34,
"learning_rate": 3.1051282051282055e-06,
"loss": 0.0001,
"step": 13950
},
{
"epoch": 81.49,
"learning_rate": 3.0923076923076927e-06,
"loss": 0.0002,
"step": 13975
},
{
"epoch": 81.63,
"learning_rate": 3.07948717948718e-06,
"loss": 0.0002,
"step": 14000
},
{
"epoch": 81.63,
"eval_loss": 0.30844250321388245,
"eval_runtime": 1315.9637,
"eval_samples_per_second": 4.861,
"eval_steps_per_second": 0.304,
"eval_wer": 6.742641718258223,
"step": 14000
},
{
"epoch": 81.78,
"learning_rate": 3.066666666666667e-06,
"loss": 0.0004,
"step": 14025
},
{
"epoch": 81.92,
"learning_rate": 3.053846153846154e-06,
"loss": 0.0003,
"step": 14050
},
{
"epoch": 82.07,
"learning_rate": 3.0410256410256413e-06,
"loss": 0.0001,
"step": 14075
},
{
"epoch": 82.22,
"learning_rate": 3.0282051282051286e-06,
"loss": 0.0001,
"step": 14100
},
{
"epoch": 82.36,
"learning_rate": 3.0153846153846154e-06,
"loss": 0.0001,
"step": 14125
},
{
"epoch": 82.51,
"learning_rate": 3.0025641025641026e-06,
"loss": 0.0001,
"step": 14150
},
{
"epoch": 82.65,
"learning_rate": 2.98974358974359e-06,
"loss": 0.0001,
"step": 14175
},
{
"epoch": 82.8,
"learning_rate": 2.976923076923077e-06,
"loss": 0.0001,
"step": 14200
},
{
"epoch": 82.94,
"learning_rate": 2.9641025641025644e-06,
"loss": 0.0003,
"step": 14225
},
{
"epoch": 83.09,
"learning_rate": 2.951794871794872e-06,
"loss": 0.0005,
"step": 14250
},
{
"epoch": 83.24,
"learning_rate": 2.938974358974359e-06,
"loss": 0.0002,
"step": 14275
},
{
"epoch": 83.38,
"learning_rate": 2.9261538461538463e-06,
"loss": 0.0002,
"step": 14300
},
{
"epoch": 83.53,
"learning_rate": 2.9133333333333335e-06,
"loss": 0.0006,
"step": 14325
},
{
"epoch": 83.67,
"learning_rate": 2.9005128205128208e-06,
"loss": 0.0002,
"step": 14350
},
{
"epoch": 83.82,
"learning_rate": 2.887692307692308e-06,
"loss": 0.0003,
"step": 14375
},
{
"epoch": 83.97,
"learning_rate": 2.8748717948717953e-06,
"loss": 0.0002,
"step": 14400
},
{
"epoch": 84.11,
"learning_rate": 2.8620512820512825e-06,
"loss": 0.0007,
"step": 14425
},
{
"epoch": 84.26,
"learning_rate": 2.8492307692307698e-06,
"loss": 0.0006,
"step": 14450
},
{
"epoch": 84.4,
"learning_rate": 2.836410256410257e-06,
"loss": 0.0006,
"step": 14475
},
{
"epoch": 84.55,
"learning_rate": 2.8235897435897434e-06,
"loss": 0.0009,
"step": 14500
},
{
"epoch": 84.69,
"learning_rate": 2.8107692307692307e-06,
"loss": 0.0002,
"step": 14525
},
{
"epoch": 84.84,
"learning_rate": 2.797948717948718e-06,
"loss": 0.0004,
"step": 14550
},
{
"epoch": 84.99,
"learning_rate": 2.785128205128205e-06,
"loss": 0.001,
"step": 14575
},
{
"epoch": 85.13,
"learning_rate": 2.7723076923076924e-06,
"loss": 0.0002,
"step": 14600
},
{
"epoch": 85.28,
"learning_rate": 2.7594871794871797e-06,
"loss": 0.0003,
"step": 14625
},
{
"epoch": 85.42,
"learning_rate": 2.746666666666667e-06,
"loss": 0.0004,
"step": 14650
},
{
"epoch": 85.57,
"learning_rate": 2.733846153846154e-06,
"loss": 0.0006,
"step": 14675
},
{
"epoch": 85.71,
"learning_rate": 2.7210256410256414e-06,
"loss": 0.0004,
"step": 14700
},
{
"epoch": 85.86,
"learning_rate": 2.7082051282051287e-06,
"loss": 0.0004,
"step": 14725
},
{
"epoch": 86.01,
"learning_rate": 2.695384615384616e-06,
"loss": 0.0002,
"step": 14750
},
{
"epoch": 86.15,
"learning_rate": 2.682564102564103e-06,
"loss": 0.0002,
"step": 14775
},
{
"epoch": 86.3,
"learning_rate": 2.6697435897435896e-06,
"loss": 0.0002,
"step": 14800
},
{
"epoch": 86.44,
"learning_rate": 2.656923076923077e-06,
"loss": 0.0005,
"step": 14825
},
{
"epoch": 86.59,
"learning_rate": 2.644102564102564e-06,
"loss": 0.0002,
"step": 14850
},
{
"epoch": 86.73,
"learning_rate": 2.6312820512820513e-06,
"loss": 0.0002,
"step": 14875
},
{
"epoch": 86.88,
"learning_rate": 2.6184615384615385e-06,
"loss": 0.0009,
"step": 14900
},
{
"epoch": 87.03,
"learning_rate": 2.605641025641026e-06,
"loss": 0.0002,
"step": 14925
},
{
"epoch": 87.17,
"learning_rate": 2.592820512820513e-06,
"loss": 0.0005,
"step": 14950
},
{
"epoch": 87.32,
"learning_rate": 2.5800000000000003e-06,
"loss": 0.0002,
"step": 14975
},
{
"epoch": 87.46,
"learning_rate": 2.5671794871794875e-06,
"loss": 0.0001,
"step": 15000
},
{
"epoch": 87.46,
"eval_loss": 0.3144780099391937,
"eval_runtime": 1315.5703,
"eval_samples_per_second": 4.863,
"eval_steps_per_second": 0.304,
"eval_wer": 6.663092726455227,
"step": 15000
},
{
"epoch": 87.61,
"learning_rate": 2.5543589743589748e-06,
"loss": 0.0001,
"step": 15025
},
{
"epoch": 87.76,
"learning_rate": 2.541538461538462e-06,
"loss": 0.0003,
"step": 15050
},
{
"epoch": 87.9,
"learning_rate": 2.5287179487179493e-06,
"loss": 0.0003,
"step": 15075
},
{
"epoch": 88.05,
"learning_rate": 2.5158974358974357e-06,
"loss": 0.0003,
"step": 15100
},
{
"epoch": 88.19,
"learning_rate": 2.503076923076923e-06,
"loss": 0.0005,
"step": 15125
},
{
"epoch": 88.34,
"learning_rate": 2.4902564102564106e-06,
"loss": 0.0006,
"step": 15150
},
{
"epoch": 88.48,
"learning_rate": 2.4774358974358974e-06,
"loss": 0.0003,
"step": 15175
},
{
"epoch": 88.63,
"learning_rate": 2.4646153846153847e-06,
"loss": 0.0011,
"step": 15200
},
{
"epoch": 88.78,
"learning_rate": 2.451794871794872e-06,
"loss": 0.0003,
"step": 15225
},
{
"epoch": 88.92,
"learning_rate": 2.438974358974359e-06,
"loss": 0.0001,
"step": 15250
},
{
"epoch": 89.07,
"learning_rate": 2.4261538461538464e-06,
"loss": 0.0003,
"step": 15275
},
{
"epoch": 89.21,
"learning_rate": 2.4133333333333337e-06,
"loss": 0.0005,
"step": 15300
},
{
"epoch": 89.36,
"learning_rate": 2.4005128205128205e-06,
"loss": 0.0002,
"step": 15325
},
{
"epoch": 89.5,
"learning_rate": 2.3876923076923077e-06,
"loss": 0.0002,
"step": 15350
},
{
"epoch": 89.65,
"learning_rate": 2.374871794871795e-06,
"loss": 0.0001,
"step": 15375
},
{
"epoch": 89.8,
"learning_rate": 2.3620512820512822e-06,
"loss": 0.0002,
"step": 15400
},
{
"epoch": 89.94,
"learning_rate": 2.3492307692307695e-06,
"loss": 0.0005,
"step": 15425
},
{
"epoch": 90.09,
"learning_rate": 2.3364102564102567e-06,
"loss": 0.0005,
"step": 15450
},
{
"epoch": 90.23,
"learning_rate": 2.3235897435897436e-06,
"loss": 0.0002,
"step": 15475
},
{
"epoch": 90.38,
"learning_rate": 2.310769230769231e-06,
"loss": 0.0001,
"step": 15500
},
{
"epoch": 90.52,
"learning_rate": 2.297948717948718e-06,
"loss": 0.0001,
"step": 15525
},
{
"epoch": 90.67,
"learning_rate": 2.2851282051282053e-06,
"loss": 0.0001,
"step": 15550
},
{
"epoch": 90.82,
"learning_rate": 2.2723076923076925e-06,
"loss": 0.0001,
"step": 15575
},
{
"epoch": 90.96,
"learning_rate": 2.25948717948718e-06,
"loss": 0.0004,
"step": 15600
},
{
"epoch": 91.11,
"learning_rate": 2.2466666666666666e-06,
"loss": 0.0001,
"step": 15625
},
{
"epoch": 91.25,
"learning_rate": 2.233846153846154e-06,
"loss": 0.0001,
"step": 15650
},
{
"epoch": 91.4,
"learning_rate": 2.221025641025641e-06,
"loss": 0.0003,
"step": 15675
},
{
"epoch": 91.55,
"learning_rate": 2.2082051282051284e-06,
"loss": 0.0001,
"step": 15700
},
{
"epoch": 91.69,
"learning_rate": 2.1953846153846156e-06,
"loss": 0.0004,
"step": 15725
},
{
"epoch": 91.84,
"learning_rate": 2.182564102564103e-06,
"loss": 0.0003,
"step": 15750
},
{
"epoch": 91.98,
"learning_rate": 2.1697435897435897e-06,
"loss": 0.0003,
"step": 15775
},
{
"epoch": 92.13,
"learning_rate": 2.156923076923077e-06,
"loss": 0.0001,
"step": 15800
},
{
"epoch": 92.27,
"learning_rate": 2.144102564102564e-06,
"loss": 0.0001,
"step": 15825
},
{
"epoch": 92.42,
"learning_rate": 2.1312820512820514e-06,
"loss": 0.0001,
"step": 15850
},
{
"epoch": 92.57,
"learning_rate": 2.1184615384615387e-06,
"loss": 0.0001,
"step": 15875
},
{
"epoch": 92.71,
"learning_rate": 2.105641025641026e-06,
"loss": 0.0002,
"step": 15900
},
{
"epoch": 92.86,
"learning_rate": 2.092820512820513e-06,
"loss": 0.0008,
"step": 15925
},
{
"epoch": 93.0,
"learning_rate": 2.08e-06,
"loss": 0.0008,
"step": 15950
},
{
"epoch": 93.15,
"learning_rate": 2.0671794871794872e-06,
"loss": 0.0009,
"step": 15975
},
{
"epoch": 93.29,
"learning_rate": 2.0543589743589745e-06,
"loss": 0.0002,
"step": 16000
},
{
"epoch": 93.29,
"eval_loss": 0.30914363265037537,
"eval_runtime": 1314.8431,
"eval_samples_per_second": 4.865,
"eval_steps_per_second": 0.304,
"eval_wer": 6.666551378272749,
"step": 16000
},
{
"epoch": 93.44,
"learning_rate": 2.0415384615384617e-06,
"loss": 0.0002,
"step": 16025
},
{
"epoch": 93.59,
"learning_rate": 2.028717948717949e-06,
"loss": 0.0001,
"step": 16050
},
{
"epoch": 93.73,
"learning_rate": 2.0158974358974362e-06,
"loss": 0.0001,
"step": 16075
},
{
"epoch": 93.88,
"learning_rate": 2.003076923076923e-06,
"loss": 0.0006,
"step": 16100
},
{
"epoch": 94.02,
"learning_rate": 1.9902564102564103e-06,
"loss": 0.0003,
"step": 16125
},
{
"epoch": 94.17,
"learning_rate": 1.9774358974358976e-06,
"loss": 0.0006,
"step": 16150
},
{
"epoch": 94.31,
"learning_rate": 1.964615384615385e-06,
"loss": 0.0001,
"step": 16175
},
{
"epoch": 94.46,
"learning_rate": 1.951794871794872e-06,
"loss": 0.0001,
"step": 16200
},
{
"epoch": 94.61,
"learning_rate": 1.9389743589743593e-06,
"loss": 0.0001,
"step": 16225
},
{
"epoch": 94.75,
"learning_rate": 1.926153846153846e-06,
"loss": 0.0001,
"step": 16250
},
{
"epoch": 94.9,
"learning_rate": 1.9133333333333334e-06,
"loss": 0.0005,
"step": 16275
},
{
"epoch": 95.04,
"learning_rate": 1.9005128205128206e-06,
"loss": 0.0001,
"step": 16300
},
{
"epoch": 95.19,
"learning_rate": 1.8876923076923079e-06,
"loss": 0.0001,
"step": 16325
},
{
"epoch": 95.34,
"learning_rate": 1.8748717948717951e-06,
"loss": 0.0001,
"step": 16350
},
{
"epoch": 95.48,
"learning_rate": 1.8620512820512824e-06,
"loss": 0.0001,
"step": 16375
},
{
"epoch": 95.63,
"learning_rate": 1.8492307692307692e-06,
"loss": 0.0004,
"step": 16400
},
{
"epoch": 95.77,
"learning_rate": 1.8364102564102564e-06,
"loss": 0.0002,
"step": 16425
},
{
"epoch": 95.92,
"learning_rate": 1.8235897435897437e-06,
"loss": 0.0002,
"step": 16450
},
{
"epoch": 96.06,
"learning_rate": 1.810769230769231e-06,
"loss": 0.0001,
"step": 16475
},
{
"epoch": 96.21,
"learning_rate": 1.7979487179487182e-06,
"loss": 0.0001,
"step": 16500
},
{
"epoch": 96.36,
"learning_rate": 1.7851282051282054e-06,
"loss": 0.0001,
"step": 16525
},
{
"epoch": 96.5,
"learning_rate": 1.7723076923076922e-06,
"loss": 0.0001,
"step": 16550
},
{
"epoch": 96.65,
"learning_rate": 1.7594871794871795e-06,
"loss": 0.0001,
"step": 16575
},
{
"epoch": 96.79,
"learning_rate": 1.7466666666666667e-06,
"loss": 0.0002,
"step": 16600
},
{
"epoch": 96.94,
"learning_rate": 1.733846153846154e-06,
"loss": 0.0001,
"step": 16625
},
{
"epoch": 97.08,
"learning_rate": 1.7210256410256412e-06,
"loss": 0.0001,
"step": 16650
},
{
"epoch": 97.23,
"learning_rate": 1.7082051282051285e-06,
"loss": 0.0001,
"step": 16675
},
{
"epoch": 97.38,
"learning_rate": 1.6953846153846153e-06,
"loss": 0.0002,
"step": 16700
},
{
"epoch": 97.52,
"learning_rate": 1.6825641025641026e-06,
"loss": 0.0001,
"step": 16725
},
{
"epoch": 97.67,
"learning_rate": 1.6697435897435898e-06,
"loss": 0.0001,
"step": 16750
},
{
"epoch": 97.81,
"learning_rate": 1.656923076923077e-06,
"loss": 0.0001,
"step": 16775
},
{
"epoch": 97.96,
"learning_rate": 1.6441025641025643e-06,
"loss": 0.0002,
"step": 16800
},
{
"epoch": 98.1,
"learning_rate": 1.6312820512820516e-06,
"loss": 0.0001,
"step": 16825
},
{
"epoch": 98.25,
"learning_rate": 1.6184615384615384e-06,
"loss": 0.0001,
"step": 16850
},
{
"epoch": 98.4,
"learning_rate": 1.6056410256410256e-06,
"loss": 0.0001,
"step": 16875
},
{
"epoch": 98.54,
"learning_rate": 1.5928205128205129e-06,
"loss": 0.0001,
"step": 16900
},
{
"epoch": 98.69,
"learning_rate": 1.5800000000000001e-06,
"loss": 0.0001,
"step": 16925
},
{
"epoch": 98.83,
"learning_rate": 1.5671794871794874e-06,
"loss": 0.0001,
"step": 16950
},
{
"epoch": 98.98,
"learning_rate": 1.5543589743589746e-06,
"loss": 0.0001,
"step": 16975
},
{
"epoch": 99.13,
"learning_rate": 1.5415384615384614e-06,
"loss": 0.0001,
"step": 17000
},
{
"epoch": 99.13,
"eval_loss": 0.31698107719421387,
"eval_runtime": 1323.9469,
"eval_samples_per_second": 4.832,
"eval_steps_per_second": 0.302,
"eval_wer": 6.875799813232803,
"step": 17000
},
{
"epoch": 99.27,
"learning_rate": 1.5287179487179487e-06,
"loss": 0.0001,
"step": 17025
},
{
"epoch": 99.42,
"learning_rate": 1.515897435897436e-06,
"loss": 0.0001,
"step": 17050
},
{
"epoch": 99.56,
"learning_rate": 1.5030769230769232e-06,
"loss": 0.0001,
"step": 17075
},
{
"epoch": 99.71,
"learning_rate": 1.4902564102564104e-06,
"loss": 0.0001,
"step": 17100
},
{
"epoch": 99.85,
"learning_rate": 1.4774358974358977e-06,
"loss": 0.0001,
"step": 17125
},
{
"epoch": 100.0,
"learning_rate": 1.4646153846153845e-06,
"loss": 0.0001,
"step": 17150
},
{
"epoch": 100.15,
"learning_rate": 1.4517948717948718e-06,
"loss": 0.0002,
"step": 17175
},
{
"epoch": 100.29,
"learning_rate": 1.438974358974359e-06,
"loss": 0.0001,
"step": 17200
},
{
"epoch": 100.44,
"learning_rate": 1.4261538461538462e-06,
"loss": 0.0001,
"step": 17225
},
{
"epoch": 100.58,
"learning_rate": 1.4133333333333335e-06,
"loss": 0.0001,
"step": 17250
},
{
"epoch": 100.73,
"learning_rate": 1.4005128205128207e-06,
"loss": 0.0001,
"step": 17275
},
{
"epoch": 100.87,
"learning_rate": 1.3876923076923076e-06,
"loss": 0.0001,
"step": 17300
},
{
"epoch": 101.02,
"learning_rate": 1.3748717948717948e-06,
"loss": 0.0001,
"step": 17325
},
{
"epoch": 101.17,
"learning_rate": 1.362051282051282e-06,
"loss": 0.0001,
"step": 17350
},
{
"epoch": 101.31,
"learning_rate": 1.3492307692307693e-06,
"loss": 0.0001,
"step": 17375
},
{
"epoch": 101.46,
"learning_rate": 1.3364102564102566e-06,
"loss": 0.0001,
"step": 17400
},
{
"epoch": 101.6,
"learning_rate": 1.3235897435897438e-06,
"loss": 0.0001,
"step": 17425
},
{
"epoch": 101.75,
"learning_rate": 1.3107692307692308e-06,
"loss": 0.0001,
"step": 17450
},
{
"epoch": 101.9,
"learning_rate": 1.2979487179487179e-06,
"loss": 0.0001,
"step": 17475
},
{
"epoch": 102.04,
"learning_rate": 1.2851282051282051e-06,
"loss": 0.0001,
"step": 17500
},
{
"epoch": 102.19,
"learning_rate": 1.2723076923076924e-06,
"loss": 0.0001,
"step": 17525
},
{
"epoch": 102.33,
"learning_rate": 1.2594871794871796e-06,
"loss": 0.0001,
"step": 17550
},
{
"epoch": 102.48,
"learning_rate": 1.2466666666666667e-06,
"loss": 0.0001,
"step": 17575
},
{
"epoch": 102.62,
"learning_rate": 1.233846153846154e-06,
"loss": 0.0001,
"step": 17600
},
{
"epoch": 102.77,
"learning_rate": 1.2210256410256412e-06,
"loss": 0.0001,
"step": 17625
},
{
"epoch": 102.92,
"learning_rate": 1.2082051282051282e-06,
"loss": 0.0001,
"step": 17650
},
{
"epoch": 103.06,
"learning_rate": 1.1953846153846154e-06,
"loss": 0.0001,
"step": 17675
},
{
"epoch": 103.21,
"learning_rate": 1.1825641025641027e-06,
"loss": 0.0002,
"step": 17700
},
{
"epoch": 103.35,
"learning_rate": 1.1697435897435897e-06,
"loss": 0.0001,
"step": 17725
},
{
"epoch": 103.5,
"learning_rate": 1.156923076923077e-06,
"loss": 0.0002,
"step": 17750
},
{
"epoch": 103.64,
"learning_rate": 1.1441025641025642e-06,
"loss": 0.0001,
"step": 17775
},
{
"epoch": 103.79,
"learning_rate": 1.1312820512820513e-06,
"loss": 0.0001,
"step": 17800
},
{
"epoch": 103.94,
"learning_rate": 1.1184615384615385e-06,
"loss": 0.0001,
"step": 17825
},
{
"epoch": 104.08,
"learning_rate": 1.1056410256410258e-06,
"loss": 0.0001,
"step": 17850
},
{
"epoch": 104.23,
"learning_rate": 1.0928205128205128e-06,
"loss": 0.0001,
"step": 17875
},
{
"epoch": 104.37,
"learning_rate": 1.08e-06,
"loss": 0.0001,
"step": 17900
},
{
"epoch": 104.52,
"learning_rate": 1.0671794871794873e-06,
"loss": 0.0001,
"step": 17925
},
{
"epoch": 104.66,
"learning_rate": 1.0543589743589743e-06,
"loss": 0.0001,
"step": 17950
},
{
"epoch": 104.81,
"learning_rate": 1.0415384615384616e-06,
"loss": 0.0001,
"step": 17975
},
{
"epoch": 104.96,
"learning_rate": 1.0287179487179488e-06,
"loss": 0.0002,
"step": 18000
},
{
"epoch": 104.96,
"eval_loss": 0.32231080532073975,
"eval_runtime": 1315.215,
"eval_samples_per_second": 4.864,
"eval_steps_per_second": 0.304,
"eval_wer": 6.633694186006295,
"step": 18000
},
{
"epoch": 105.1,
"learning_rate": 1.0158974358974359e-06,
"loss": 0.0002,
"step": 18025
},
{
"epoch": 105.25,
"learning_rate": 1.003076923076923e-06,
"loss": 0.0001,
"step": 18050
},
{
"epoch": 105.39,
"learning_rate": 9.902564102564103e-07,
"loss": 0.0001,
"step": 18075
},
{
"epoch": 105.54,
"learning_rate": 9.774358974358974e-07,
"loss": 0.0001,
"step": 18100
},
{
"epoch": 105.69,
"learning_rate": 9.646153846153846e-07,
"loss": 0.0002,
"step": 18125
},
{
"epoch": 105.83,
"learning_rate": 9.517948717948719e-07,
"loss": 0.0001,
"step": 18150
},
{
"epoch": 105.98,
"learning_rate": 9.38974358974359e-07,
"loss": 0.0001,
"step": 18175
},
{
"epoch": 106.12,
"learning_rate": 9.261538461538462e-07,
"loss": 0.0001,
"step": 18200
},
{
"epoch": 106.27,
"learning_rate": 9.133333333333334e-07,
"loss": 0.0001,
"step": 18225
},
{
"epoch": 106.41,
"learning_rate": 9.005128205128206e-07,
"loss": 0.0001,
"step": 18250
},
{
"epoch": 106.56,
"learning_rate": 8.876923076923077e-07,
"loss": 0.0001,
"step": 18275
},
{
"epoch": 106.71,
"learning_rate": 8.748717948717949e-07,
"loss": 0.0001,
"step": 18300
},
{
"epoch": 106.85,
"learning_rate": 8.620512820512822e-07,
"loss": 0.0001,
"step": 18325
},
{
"epoch": 107.0,
"learning_rate": 8.492307692307692e-07,
"loss": 0.0001,
"step": 18350
},
{
"epoch": 107.14,
"learning_rate": 8.364102564102565e-07,
"loss": 0.0004,
"step": 18375
},
{
"epoch": 107.29,
"learning_rate": 8.235897435897437e-07,
"loss": 0.0001,
"step": 18400
},
{
"epoch": 107.43,
"learning_rate": 8.107692307692308e-07,
"loss": 0.0001,
"step": 18425
},
{
"epoch": 107.58,
"learning_rate": 7.97948717948718e-07,
"loss": 0.0001,
"step": 18450
},
{
"epoch": 107.73,
"learning_rate": 7.851282051282053e-07,
"loss": 0.0001,
"step": 18475
},
{
"epoch": 107.87,
"learning_rate": 7.723076923076923e-07,
"loss": 0.0001,
"step": 18500
},
{
"epoch": 108.02,
"learning_rate": 7.594871794871795e-07,
"loss": 0.0001,
"step": 18525
},
{
"epoch": 108.16,
"learning_rate": 7.466666666666668e-07,
"loss": 0.0001,
"step": 18550
},
{
"epoch": 108.31,
"learning_rate": 7.338461538461538e-07,
"loss": 0.0001,
"step": 18575
},
{
"epoch": 108.45,
"learning_rate": 7.210256410256411e-07,
"loss": 0.0001,
"step": 18600
},
{
"epoch": 108.6,
"learning_rate": 7.082051282051283e-07,
"loss": 0.0001,
"step": 18625
},
{
"epoch": 108.75,
"learning_rate": 6.953846153846154e-07,
"loss": 0.0001,
"step": 18650
},
{
"epoch": 108.89,
"learning_rate": 6.825641025641026e-07,
"loss": 0.0001,
"step": 18675
},
{
"epoch": 109.04,
"learning_rate": 6.697435897435899e-07,
"loss": 0.0001,
"step": 18700
},
{
"epoch": 109.18,
"learning_rate": 6.569230769230769e-07,
"loss": 0.0001,
"step": 18725
},
{
"epoch": 109.33,
"learning_rate": 6.441025641025641e-07,
"loss": 0.0001,
"step": 18750
},
{
"epoch": 109.48,
"learning_rate": 6.312820512820514e-07,
"loss": 0.0001,
"step": 18775
},
{
"epoch": 109.62,
"learning_rate": 6.184615384615385e-07,
"loss": 0.0001,
"step": 18800
},
{
"epoch": 109.77,
"learning_rate": 6.056410256410257e-07,
"loss": 0.0,
"step": 18825
},
{
"epoch": 109.91,
"learning_rate": 5.928205128205128e-07,
"loss": 0.0,
"step": 18850
},
{
"epoch": 110.06,
"learning_rate": 5.800000000000001e-07,
"loss": 0.0002,
"step": 18875
},
{
"epoch": 110.2,
"learning_rate": 5.671794871794872e-07,
"loss": 0.0001,
"step": 18900
},
{
"epoch": 110.35,
"learning_rate": 5.543589743589743e-07,
"loss": 0.0001,
"step": 18925
},
{
"epoch": 110.5,
"learning_rate": 5.415384615384616e-07,
"loss": 0.0001,
"step": 18950
},
{
"epoch": 110.64,
"learning_rate": 5.287179487179487e-07,
"loss": 0.0001,
"step": 18975
},
{
"epoch": 110.79,
"learning_rate": 5.158974358974359e-07,
"loss": 0.0,
"step": 19000
},
{
"epoch": 110.79,
"eval_loss": 0.32186684012413025,
"eval_runtime": 1314.7171,
"eval_samples_per_second": 4.866,
"eval_steps_per_second": 0.304,
"eval_wer": 6.497077439214194,
"step": 19000
},
{
"epoch": 110.93,
"learning_rate": 5.030769230769231e-07,
"loss": 0.0003,
"step": 19025
},
{
"epoch": 111.08,
"learning_rate": 4.902564102564103e-07,
"loss": 0.0001,
"step": 19050
},
{
"epoch": 111.22,
"learning_rate": 4.774358974358974e-07,
"loss": 0.0001,
"step": 19075
},
{
"epoch": 111.37,
"learning_rate": 4.6461538461538465e-07,
"loss": 0.0,
"step": 19100
},
{
"epoch": 111.52,
"learning_rate": 4.5179487179487185e-07,
"loss": 0.0001,
"step": 19125
},
{
"epoch": 111.66,
"learning_rate": 4.38974358974359e-07,
"loss": 0.0002,
"step": 19150
},
{
"epoch": 111.81,
"learning_rate": 4.261538461538462e-07,
"loss": 0.0001,
"step": 19175
},
{
"epoch": 111.95,
"learning_rate": 4.133333333333334e-07,
"loss": 0.0,
"step": 19200
},
{
"epoch": 112.1,
"learning_rate": 4.005128205128205e-07,
"loss": 0.0001,
"step": 19225
},
{
"epoch": 112.24,
"learning_rate": 3.876923076923077e-07,
"loss": 0.0,
"step": 19250
},
{
"epoch": 112.39,
"learning_rate": 3.748717948717949e-07,
"loss": 0.0,
"step": 19275
},
{
"epoch": 112.54,
"learning_rate": 3.6205128205128206e-07,
"loss": 0.0,
"step": 19300
},
{
"epoch": 112.68,
"learning_rate": 3.4923076923076925e-07,
"loss": 0.0002,
"step": 19325
},
{
"epoch": 112.83,
"learning_rate": 3.3641025641025645e-07,
"loss": 0.0001,
"step": 19350
},
{
"epoch": 112.97,
"learning_rate": 3.235897435897436e-07,
"loss": 0.0,
"step": 19375
},
{
"epoch": 113.12,
"learning_rate": 3.107692307692308e-07,
"loss": 0.0001,
"step": 19400
},
{
"epoch": 113.27,
"learning_rate": 2.97948717948718e-07,
"loss": 0.0,
"step": 19425
},
{
"epoch": 113.41,
"learning_rate": 2.8512820512820517e-07,
"loss": 0.0001,
"step": 19450
},
{
"epoch": 113.56,
"learning_rate": 2.7230769230769237e-07,
"loss": 0.0001,
"step": 19475
},
{
"epoch": 113.7,
"learning_rate": 2.594871794871795e-07,
"loss": 0.0001,
"step": 19500
},
{
"epoch": 113.85,
"learning_rate": 2.466666666666667e-07,
"loss": 0.0,
"step": 19525
},
{
"epoch": 113.99,
"learning_rate": 2.3384615384615387e-07,
"loss": 0.0,
"step": 19550
},
{
"epoch": 114.14,
"learning_rate": 2.2102564102564104e-07,
"loss": 0.0001,
"step": 19575
},
{
"epoch": 114.29,
"learning_rate": 2.082051282051282e-07,
"loss": 0.0,
"step": 19600
},
{
"epoch": 114.43,
"learning_rate": 1.953846153846154e-07,
"loss": 0.0,
"step": 19625
},
{
"epoch": 114.58,
"learning_rate": 1.8256410256410257e-07,
"loss": 0.0001,
"step": 19650
},
{
"epoch": 114.72,
"learning_rate": 1.6974358974358974e-07,
"loss": 0.0,
"step": 19675
},
{
"epoch": 114.87,
"learning_rate": 1.5692307692307694e-07,
"loss": 0.0001,
"step": 19700
},
{
"epoch": 115.01,
"learning_rate": 1.441025641025641e-07,
"loss": 0.0,
"step": 19725
},
{
"epoch": 115.16,
"learning_rate": 1.3128205128205127e-07,
"loss": 0.0,
"step": 19750
},
{
"epoch": 115.31,
"learning_rate": 1.1846153846153847e-07,
"loss": 0.0,
"step": 19775
},
{
"epoch": 115.45,
"learning_rate": 1.0564102564102565e-07,
"loss": 0.0,
"step": 19800
},
{
"epoch": 115.6,
"learning_rate": 9.282051282051283e-08,
"loss": 0.0001,
"step": 19825
},
{
"epoch": 115.74,
"learning_rate": 8e-08,
"loss": 0.0001,
"step": 19850
},
{
"epoch": 115.89,
"learning_rate": 6.717948717948718e-08,
"loss": 0.0,
"step": 19875
},
{
"epoch": 116.03,
"learning_rate": 5.435897435897436e-08,
"loss": 0.0001,
"step": 19900
},
{
"epoch": 116.18,
"learning_rate": 4.153846153846154e-08,
"loss": 0.0001,
"step": 19925
},
{
"epoch": 116.33,
"learning_rate": 2.8717948717948722e-08,
"loss": 0.0,
"step": 19950
},
{
"epoch": 116.47,
"learning_rate": 1.5897435897435897e-08,
"loss": 0.0,
"step": 19975
},
{
"epoch": 116.62,
"learning_rate": 3.0769230769230774e-09,
"loss": 0.0001,
"step": 20000
},
{
"epoch": 116.62,
"eval_loss": 0.3219425082206726,
"eval_runtime": 1315.0644,
"eval_samples_per_second": 4.864,
"eval_steps_per_second": 0.304,
"eval_wer": 6.500536091031715,
"step": 20000
},
{
"epoch": 116.62,
"step": 20000,
"total_flos": 2.7115017106636716e+21,
"train_loss": 0.017075960696069523,
"train_runtime": 139630.5899,
"train_samples_per_second": 9.167,
"train_steps_per_second": 0.143
}
],
"logging_steps": 25,
"max_steps": 20000,
"num_input_tokens_seen": 0,
"num_train_epochs": 117,
"save_steps": 1000,
"total_flos": 2.7115017106636716e+21,
"train_batch_size": 32,
"trial_name": null,
"trial_params": null
}