whisper-medium-mn-4 / trainer_state.json
bayartsogt's picture
End of training
8b0ea03
{
"best_metric": 33.029276818876994,
"best_model_checkpoint": "whisper-medium-mn-4/checkpoint-14000",
"epoch": 63.829787234042556,
"global_step": 15000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.11,
"learning_rate": 4.2000000000000006e-07,
"loss": 4.4672,
"step": 25
},
{
"epoch": 0.21,
"learning_rate": 9.200000000000001e-07,
"loss": 3.515,
"step": 50
},
{
"epoch": 0.32,
"learning_rate": 1.42e-06,
"loss": 2.5788,
"step": 75
},
{
"epoch": 0.43,
"learning_rate": 1.9200000000000003e-06,
"loss": 1.9034,
"step": 100
},
{
"epoch": 0.53,
"learning_rate": 2.42e-06,
"loss": 1.4762,
"step": 125
},
{
"epoch": 0.64,
"learning_rate": 2.92e-06,
"loss": 1.1463,
"step": 150
},
{
"epoch": 0.74,
"learning_rate": 3.4200000000000007e-06,
"loss": 0.8071,
"step": 175
},
{
"epoch": 0.85,
"learning_rate": 3.920000000000001e-06,
"loss": 0.6609,
"step": 200
},
{
"epoch": 0.96,
"learning_rate": 4.42e-06,
"loss": 0.5819,
"step": 225
},
{
"epoch": 1.06,
"learning_rate": 4.92e-06,
"loss": 0.5074,
"step": 250
},
{
"epoch": 1.17,
"learning_rate": 5.420000000000001e-06,
"loss": 0.441,
"step": 275
},
{
"epoch": 1.28,
"learning_rate": 5.92e-06,
"loss": 0.3912,
"step": 300
},
{
"epoch": 1.38,
"learning_rate": 6.42e-06,
"loss": 0.3785,
"step": 325
},
{
"epoch": 1.49,
"learning_rate": 6.92e-06,
"loss": 0.3495,
"step": 350
},
{
"epoch": 1.6,
"learning_rate": 7.420000000000001e-06,
"loss": 0.3382,
"step": 375
},
{
"epoch": 1.7,
"learning_rate": 7.92e-06,
"loss": 0.3386,
"step": 400
},
{
"epoch": 1.81,
"learning_rate": 8.42e-06,
"loss": 0.3001,
"step": 425
},
{
"epoch": 1.91,
"learning_rate": 8.920000000000001e-06,
"loss": 0.2959,
"step": 450
},
{
"epoch": 2.02,
"learning_rate": 9.42e-06,
"loss": 0.2704,
"step": 475
},
{
"epoch": 2.13,
"learning_rate": 9.920000000000002e-06,
"loss": 0.1755,
"step": 500
},
{
"epoch": 2.23,
"learning_rate": 9.98551724137931e-06,
"loss": 0.1838,
"step": 525
},
{
"epoch": 2.34,
"learning_rate": 9.968275862068967e-06,
"loss": 0.1771,
"step": 550
},
{
"epoch": 2.45,
"learning_rate": 9.951034482758622e-06,
"loss": 0.1697,
"step": 575
},
{
"epoch": 2.55,
"learning_rate": 9.933793103448277e-06,
"loss": 0.1661,
"step": 600
},
{
"epoch": 2.66,
"learning_rate": 9.916551724137932e-06,
"loss": 0.1715,
"step": 625
},
{
"epoch": 2.77,
"learning_rate": 9.899310344827588e-06,
"loss": 0.1713,
"step": 650
},
{
"epoch": 2.87,
"learning_rate": 9.882068965517241e-06,
"loss": 0.1577,
"step": 675
},
{
"epoch": 2.98,
"learning_rate": 9.864827586206898e-06,
"loss": 0.1514,
"step": 700
},
{
"epoch": 3.09,
"learning_rate": 9.847586206896553e-06,
"loss": 0.0839,
"step": 725
},
{
"epoch": 3.19,
"learning_rate": 9.830344827586208e-06,
"loss": 0.0764,
"step": 750
},
{
"epoch": 3.3,
"learning_rate": 9.813103448275862e-06,
"loss": 0.0778,
"step": 775
},
{
"epoch": 3.4,
"learning_rate": 9.795862068965517e-06,
"loss": 0.071,
"step": 800
},
{
"epoch": 3.51,
"learning_rate": 9.778620689655172e-06,
"loss": 0.076,
"step": 825
},
{
"epoch": 3.62,
"learning_rate": 9.761379310344829e-06,
"loss": 0.0794,
"step": 850
},
{
"epoch": 3.72,
"learning_rate": 9.744137931034484e-06,
"loss": 0.0769,
"step": 875
},
{
"epoch": 3.83,
"learning_rate": 9.726896551724139e-06,
"loss": 0.0741,
"step": 900
},
{
"epoch": 3.94,
"learning_rate": 9.709655172413795e-06,
"loss": 0.0794,
"step": 925
},
{
"epoch": 4.04,
"learning_rate": 9.692413793103448e-06,
"loss": 0.062,
"step": 950
},
{
"epoch": 4.15,
"learning_rate": 9.675172413793105e-06,
"loss": 0.0363,
"step": 975
},
{
"epoch": 4.26,
"learning_rate": 9.65793103448276e-06,
"loss": 0.0362,
"step": 1000
},
{
"epoch": 4.26,
"eval_cer": 13.838940697614508,
"eval_loss": 0.42037829756736755,
"eval_runtime": 2299.2229,
"eval_samples_per_second": 0.82,
"eval_steps_per_second": 0.051,
"eval_wer": 40.27201223508848,
"step": 1000
},
{
"epoch": 4.36,
"learning_rate": 9.640689655172415e-06,
"loss": 0.0386,
"step": 1025
},
{
"epoch": 4.47,
"learning_rate": 9.62344827586207e-06,
"loss": 0.033,
"step": 1050
},
{
"epoch": 4.57,
"learning_rate": 9.606206896551726e-06,
"loss": 0.0406,
"step": 1075
},
{
"epoch": 4.68,
"learning_rate": 9.58896551724138e-06,
"loss": 0.0379,
"step": 1100
},
{
"epoch": 4.79,
"learning_rate": 9.571724137931036e-06,
"loss": 0.0372,
"step": 1125
},
{
"epoch": 4.89,
"learning_rate": 9.55448275862069e-06,
"loss": 0.0399,
"step": 1150
},
{
"epoch": 5.0,
"learning_rate": 9.537241379310345e-06,
"loss": 0.0378,
"step": 1175
},
{
"epoch": 5.11,
"learning_rate": 9.52e-06,
"loss": 0.0185,
"step": 1200
},
{
"epoch": 5.21,
"learning_rate": 9.502758620689655e-06,
"loss": 0.0209,
"step": 1225
},
{
"epoch": 5.32,
"learning_rate": 9.48551724137931e-06,
"loss": 0.0219,
"step": 1250
},
{
"epoch": 5.43,
"learning_rate": 9.468275862068967e-06,
"loss": 0.0201,
"step": 1275
},
{
"epoch": 5.53,
"learning_rate": 9.451034482758622e-06,
"loss": 0.0234,
"step": 1300
},
{
"epoch": 5.64,
"learning_rate": 9.433793103448276e-06,
"loss": 0.0266,
"step": 1325
},
{
"epoch": 5.74,
"learning_rate": 9.416551724137933e-06,
"loss": 0.0244,
"step": 1350
},
{
"epoch": 5.85,
"learning_rate": 9.399310344827586e-06,
"loss": 0.0227,
"step": 1375
},
{
"epoch": 5.96,
"learning_rate": 9.382068965517243e-06,
"loss": 0.0217,
"step": 1400
},
{
"epoch": 6.06,
"learning_rate": 9.364827586206898e-06,
"loss": 0.0148,
"step": 1425
},
{
"epoch": 6.17,
"learning_rate": 9.347586206896552e-06,
"loss": 0.012,
"step": 1450
},
{
"epoch": 6.28,
"learning_rate": 9.330344827586207e-06,
"loss": 0.0165,
"step": 1475
},
{
"epoch": 6.38,
"learning_rate": 9.313103448275864e-06,
"loss": 0.014,
"step": 1500
},
{
"epoch": 6.49,
"learning_rate": 9.295862068965517e-06,
"loss": 0.0146,
"step": 1525
},
{
"epoch": 6.6,
"learning_rate": 9.278620689655174e-06,
"loss": 0.0159,
"step": 1550
},
{
"epoch": 6.7,
"learning_rate": 9.261379310344828e-06,
"loss": 0.0168,
"step": 1575
},
{
"epoch": 6.81,
"learning_rate": 9.244137931034483e-06,
"loss": 0.0158,
"step": 1600
},
{
"epoch": 6.91,
"learning_rate": 9.226896551724138e-06,
"loss": 0.0147,
"step": 1625
},
{
"epoch": 7.02,
"learning_rate": 9.209655172413793e-06,
"loss": 0.0153,
"step": 1650
},
{
"epoch": 7.13,
"learning_rate": 9.192413793103448e-06,
"loss": 0.009,
"step": 1675
},
{
"epoch": 7.23,
"learning_rate": 9.175172413793105e-06,
"loss": 0.0096,
"step": 1700
},
{
"epoch": 7.34,
"learning_rate": 9.15793103448276e-06,
"loss": 0.01,
"step": 1725
},
{
"epoch": 7.45,
"learning_rate": 9.140689655172414e-06,
"loss": 0.0118,
"step": 1750
},
{
"epoch": 7.55,
"learning_rate": 9.12344827586207e-06,
"loss": 0.0103,
"step": 1775
},
{
"epoch": 7.66,
"learning_rate": 9.106206896551724e-06,
"loss": 0.0104,
"step": 1800
},
{
"epoch": 7.77,
"learning_rate": 9.08896551724138e-06,
"loss": 0.0102,
"step": 1825
},
{
"epoch": 7.87,
"learning_rate": 9.071724137931035e-06,
"loss": 0.0114,
"step": 1850
},
{
"epoch": 7.98,
"learning_rate": 9.05448275862069e-06,
"loss": 0.0109,
"step": 1875
},
{
"epoch": 8.09,
"learning_rate": 9.037241379310345e-06,
"loss": 0.0078,
"step": 1900
},
{
"epoch": 8.19,
"learning_rate": 9.020000000000002e-06,
"loss": 0.0083,
"step": 1925
},
{
"epoch": 8.3,
"learning_rate": 9.002758620689655e-06,
"loss": 0.009,
"step": 1950
},
{
"epoch": 8.4,
"learning_rate": 8.985517241379311e-06,
"loss": 0.0091,
"step": 1975
},
{
"epoch": 8.51,
"learning_rate": 8.968275862068966e-06,
"loss": 0.0087,
"step": 2000
},
{
"epoch": 8.51,
"eval_cer": 12.917546842530097,
"eval_loss": 0.47117722034454346,
"eval_runtime": 2248.6295,
"eval_samples_per_second": 0.839,
"eval_steps_per_second": 0.052,
"eval_wer": 37.4918068603889,
"step": 2000
},
{
"epoch": 8.62,
"learning_rate": 8.951034482758621e-06,
"loss": 0.0073,
"step": 2025
},
{
"epoch": 8.72,
"learning_rate": 8.933793103448276e-06,
"loss": 0.0079,
"step": 2050
},
{
"epoch": 8.83,
"learning_rate": 8.916551724137931e-06,
"loss": 0.0089,
"step": 2075
},
{
"epoch": 8.94,
"learning_rate": 8.899310344827588e-06,
"loss": 0.0089,
"step": 2100
},
{
"epoch": 9.04,
"learning_rate": 8.882068965517242e-06,
"loss": 0.008,
"step": 2125
},
{
"epoch": 9.15,
"learning_rate": 8.864827586206897e-06,
"loss": 0.0053,
"step": 2150
},
{
"epoch": 9.26,
"learning_rate": 8.847586206896552e-06,
"loss": 0.0057,
"step": 2175
},
{
"epoch": 9.36,
"learning_rate": 8.830344827586209e-06,
"loss": 0.0064,
"step": 2200
},
{
"epoch": 9.47,
"learning_rate": 8.813103448275862e-06,
"loss": 0.0066,
"step": 2225
},
{
"epoch": 9.57,
"learning_rate": 8.795862068965518e-06,
"loss": 0.0075,
"step": 2250
},
{
"epoch": 9.68,
"learning_rate": 8.778620689655173e-06,
"loss": 0.0067,
"step": 2275
},
{
"epoch": 9.79,
"learning_rate": 8.761379310344828e-06,
"loss": 0.0074,
"step": 2300
},
{
"epoch": 9.89,
"learning_rate": 8.744137931034483e-06,
"loss": 0.0056,
"step": 2325
},
{
"epoch": 10.0,
"learning_rate": 8.72689655172414e-06,
"loss": 0.0062,
"step": 2350
},
{
"epoch": 10.11,
"learning_rate": 8.709655172413793e-06,
"loss": 0.0048,
"step": 2375
},
{
"epoch": 10.21,
"learning_rate": 8.69241379310345e-06,
"loss": 0.0054,
"step": 2400
},
{
"epoch": 10.32,
"learning_rate": 8.675172413793104e-06,
"loss": 0.0048,
"step": 2425
},
{
"epoch": 10.43,
"learning_rate": 8.657931034482759e-06,
"loss": 0.0056,
"step": 2450
},
{
"epoch": 10.53,
"learning_rate": 8.640689655172414e-06,
"loss": 0.0057,
"step": 2475
},
{
"epoch": 10.64,
"learning_rate": 8.623448275862069e-06,
"loss": 0.0058,
"step": 2500
},
{
"epoch": 10.74,
"learning_rate": 8.606206896551725e-06,
"loss": 0.0065,
"step": 2525
},
{
"epoch": 10.85,
"learning_rate": 8.58896551724138e-06,
"loss": 0.0056,
"step": 2550
},
{
"epoch": 10.96,
"learning_rate": 8.571724137931035e-06,
"loss": 0.0063,
"step": 2575
},
{
"epoch": 11.06,
"learning_rate": 8.55448275862069e-06,
"loss": 0.0055,
"step": 2600
},
{
"epoch": 11.17,
"learning_rate": 8.537241379310347e-06,
"loss": 0.0048,
"step": 2625
},
{
"epoch": 11.28,
"learning_rate": 8.52e-06,
"loss": 0.0052,
"step": 2650
},
{
"epoch": 11.38,
"learning_rate": 8.502758620689656e-06,
"loss": 0.0049,
"step": 2675
},
{
"epoch": 11.49,
"learning_rate": 8.485517241379311e-06,
"loss": 0.0042,
"step": 2700
},
{
"epoch": 11.6,
"learning_rate": 8.468275862068966e-06,
"loss": 0.0069,
"step": 2725
},
{
"epoch": 11.7,
"learning_rate": 8.451034482758621e-06,
"loss": 0.0052,
"step": 2750
},
{
"epoch": 11.81,
"learning_rate": 8.433793103448277e-06,
"loss": 0.0049,
"step": 2775
},
{
"epoch": 11.91,
"learning_rate": 8.41655172413793e-06,
"loss": 0.0052,
"step": 2800
},
{
"epoch": 12.02,
"learning_rate": 8.399310344827587e-06,
"loss": 0.0056,
"step": 2825
},
{
"epoch": 12.13,
"learning_rate": 8.382068965517242e-06,
"loss": 0.0055,
"step": 2850
},
{
"epoch": 12.23,
"learning_rate": 8.364827586206897e-06,
"loss": 0.003,
"step": 2875
},
{
"epoch": 12.34,
"learning_rate": 8.347586206896552e-06,
"loss": 0.0032,
"step": 2900
},
{
"epoch": 12.45,
"learning_rate": 8.330344827586208e-06,
"loss": 0.0046,
"step": 2925
},
{
"epoch": 12.55,
"learning_rate": 8.313103448275863e-06,
"loss": 0.0036,
"step": 2950
},
{
"epoch": 12.66,
"learning_rate": 8.295862068965518e-06,
"loss": 0.0047,
"step": 2975
},
{
"epoch": 12.77,
"learning_rate": 8.278620689655173e-06,
"loss": 0.0044,
"step": 3000
},
{
"epoch": 12.77,
"eval_cer": 12.472736015937622,
"eval_loss": 0.4892961382865906,
"eval_runtime": 2214.8374,
"eval_samples_per_second": 0.852,
"eval_steps_per_second": 0.053,
"eval_wer": 36.33930522176098,
"step": 3000
},
{
"epoch": 12.87,
"learning_rate": 8.261379310344828e-06,
"loss": 0.0061,
"step": 3025
},
{
"epoch": 12.98,
"learning_rate": 8.244137931034484e-06,
"loss": 0.004,
"step": 3050
},
{
"epoch": 13.09,
"learning_rate": 8.226896551724138e-06,
"loss": 0.0032,
"step": 3075
},
{
"epoch": 13.19,
"learning_rate": 8.209655172413794e-06,
"loss": 0.0027,
"step": 3100
},
{
"epoch": 13.3,
"learning_rate": 8.192413793103449e-06,
"loss": 0.0031,
"step": 3125
},
{
"epoch": 13.4,
"learning_rate": 8.175172413793104e-06,
"loss": 0.0033,
"step": 3150
},
{
"epoch": 13.51,
"learning_rate": 8.157931034482759e-06,
"loss": 0.0035,
"step": 3175
},
{
"epoch": 13.62,
"learning_rate": 8.140689655172415e-06,
"loss": 0.0027,
"step": 3200
},
{
"epoch": 13.72,
"learning_rate": 8.123448275862069e-06,
"loss": 0.003,
"step": 3225
},
{
"epoch": 13.83,
"learning_rate": 8.106206896551725e-06,
"loss": 0.0035,
"step": 3250
},
{
"epoch": 13.94,
"learning_rate": 8.08896551724138e-06,
"loss": 0.0047,
"step": 3275
},
{
"epoch": 14.04,
"learning_rate": 8.071724137931035e-06,
"loss": 0.0032,
"step": 3300
},
{
"epoch": 14.15,
"learning_rate": 8.054482758620691e-06,
"loss": 0.0027,
"step": 3325
},
{
"epoch": 14.26,
"learning_rate": 8.037241379310346e-06,
"loss": 0.0028,
"step": 3350
},
{
"epoch": 14.36,
"learning_rate": 8.020000000000001e-06,
"loss": 0.0026,
"step": 3375
},
{
"epoch": 14.47,
"learning_rate": 8.002758620689656e-06,
"loss": 0.0038,
"step": 3400
},
{
"epoch": 14.57,
"learning_rate": 7.985517241379311e-06,
"loss": 0.0031,
"step": 3425
},
{
"epoch": 14.68,
"learning_rate": 7.968275862068966e-06,
"loss": 0.0035,
"step": 3450
},
{
"epoch": 14.79,
"learning_rate": 7.951034482758622e-06,
"loss": 0.0032,
"step": 3475
},
{
"epoch": 14.89,
"learning_rate": 7.933793103448275e-06,
"loss": 0.0039,
"step": 3500
},
{
"epoch": 15.0,
"learning_rate": 7.916551724137932e-06,
"loss": 0.0042,
"step": 3525
},
{
"epoch": 15.11,
"learning_rate": 7.899310344827587e-06,
"loss": 0.0027,
"step": 3550
},
{
"epoch": 15.21,
"learning_rate": 7.882068965517242e-06,
"loss": 0.0021,
"step": 3575
},
{
"epoch": 15.32,
"learning_rate": 7.864827586206897e-06,
"loss": 0.0024,
"step": 3600
},
{
"epoch": 15.43,
"learning_rate": 7.847586206896553e-06,
"loss": 0.003,
"step": 3625
},
{
"epoch": 15.53,
"learning_rate": 7.830344827586206e-06,
"loss": 0.0025,
"step": 3650
},
{
"epoch": 15.64,
"learning_rate": 7.813103448275863e-06,
"loss": 0.0026,
"step": 3675
},
{
"epoch": 15.74,
"learning_rate": 7.795862068965518e-06,
"loss": 0.0033,
"step": 3700
},
{
"epoch": 15.85,
"learning_rate": 7.778620689655173e-06,
"loss": 0.003,
"step": 3725
},
{
"epoch": 15.96,
"learning_rate": 7.76137931034483e-06,
"loss": 0.0032,
"step": 3750
},
{
"epoch": 16.06,
"learning_rate": 7.744137931034484e-06,
"loss": 0.0032,
"step": 3775
},
{
"epoch": 16.17,
"learning_rate": 7.726896551724139e-06,
"loss": 0.0027,
"step": 3800
},
{
"epoch": 16.28,
"learning_rate": 7.709655172413794e-06,
"loss": 0.0024,
"step": 3825
},
{
"epoch": 16.38,
"learning_rate": 7.692413793103449e-06,
"loss": 0.0023,
"step": 3850
},
{
"epoch": 16.49,
"learning_rate": 7.675172413793104e-06,
"loss": 0.0015,
"step": 3875
},
{
"epoch": 16.6,
"learning_rate": 7.65793103448276e-06,
"loss": 0.0018,
"step": 3900
},
{
"epoch": 16.7,
"learning_rate": 7.640689655172413e-06,
"loss": 0.002,
"step": 3925
},
{
"epoch": 16.81,
"learning_rate": 7.62344827586207e-06,
"loss": 0.0023,
"step": 3950
},
{
"epoch": 16.91,
"learning_rate": 7.606206896551725e-06,
"loss": 0.0027,
"step": 3975
},
{
"epoch": 17.02,
"learning_rate": 7.58896551724138e-06,
"loss": 0.0033,
"step": 4000
},
{
"epoch": 17.02,
"eval_cer": 12.293266010613634,
"eval_loss": 0.5158801674842834,
"eval_runtime": 2204.8897,
"eval_samples_per_second": 0.855,
"eval_steps_per_second": 0.054,
"eval_wer": 35.84225475202098,
"step": 4000
},
{
"epoch": 17.13,
"learning_rate": 7.5717241379310345e-06,
"loss": 0.0017,
"step": 4025
},
{
"epoch": 17.23,
"learning_rate": 7.55448275862069e-06,
"loss": 0.0025,
"step": 4050
},
{
"epoch": 17.34,
"learning_rate": 7.537241379310345e-06,
"loss": 0.003,
"step": 4075
},
{
"epoch": 17.45,
"learning_rate": 7.520000000000001e-06,
"loss": 0.0028,
"step": 4100
},
{
"epoch": 17.55,
"learning_rate": 7.5027586206896566e-06,
"loss": 0.0023,
"step": 4125
},
{
"epoch": 17.66,
"learning_rate": 7.485517241379311e-06,
"loss": 0.0029,
"step": 4150
},
{
"epoch": 17.77,
"learning_rate": 7.468275862068966e-06,
"loss": 0.0022,
"step": 4175
},
{
"epoch": 17.87,
"learning_rate": 7.451034482758621e-06,
"loss": 0.0025,
"step": 4200
},
{
"epoch": 17.98,
"learning_rate": 7.433793103448277e-06,
"loss": 0.0023,
"step": 4225
},
{
"epoch": 18.09,
"learning_rate": 7.416551724137932e-06,
"loss": 0.0019,
"step": 4250
},
{
"epoch": 18.19,
"learning_rate": 7.3993103448275875e-06,
"loss": 0.0024,
"step": 4275
},
{
"epoch": 18.3,
"learning_rate": 7.3820689655172415e-06,
"loss": 0.0023,
"step": 4300
},
{
"epoch": 18.4,
"learning_rate": 7.364827586206897e-06,
"loss": 0.0027,
"step": 4325
},
{
"epoch": 18.51,
"learning_rate": 7.347586206896552e-06,
"loss": 0.0032,
"step": 4350
},
{
"epoch": 18.62,
"learning_rate": 7.330344827586208e-06,
"loss": 0.003,
"step": 4375
},
{
"epoch": 18.72,
"learning_rate": 7.313103448275863e-06,
"loss": 0.0031,
"step": 4400
},
{
"epoch": 18.83,
"learning_rate": 7.2958620689655175e-06,
"loss": 0.0028,
"step": 4425
},
{
"epoch": 18.94,
"learning_rate": 7.278620689655172e-06,
"loss": 0.003,
"step": 4450
},
{
"epoch": 19.04,
"learning_rate": 7.261379310344828e-06,
"loss": 0.0038,
"step": 4475
},
{
"epoch": 19.15,
"learning_rate": 7.244137931034483e-06,
"loss": 0.0022,
"step": 4500
},
{
"epoch": 19.26,
"learning_rate": 7.226896551724139e-06,
"loss": 0.0028,
"step": 4525
},
{
"epoch": 19.36,
"learning_rate": 7.2096551724137944e-06,
"loss": 0.0018,
"step": 4550
},
{
"epoch": 19.47,
"learning_rate": 7.1924137931034485e-06,
"loss": 0.0023,
"step": 4575
},
{
"epoch": 19.57,
"learning_rate": 7.175172413793104e-06,
"loss": 0.0018,
"step": 4600
},
{
"epoch": 19.68,
"learning_rate": 7.157931034482759e-06,
"loss": 0.0015,
"step": 4625
},
{
"epoch": 19.79,
"learning_rate": 7.140689655172415e-06,
"loss": 0.0019,
"step": 4650
},
{
"epoch": 19.89,
"learning_rate": 7.12344827586207e-06,
"loss": 0.0026,
"step": 4675
},
{
"epoch": 20.0,
"learning_rate": 7.106206896551725e-06,
"loss": 0.003,
"step": 4700
},
{
"epoch": 20.11,
"learning_rate": 7.088965517241379e-06,
"loss": 0.0017,
"step": 4725
},
{
"epoch": 20.21,
"learning_rate": 7.071724137931035e-06,
"loss": 0.0009,
"step": 4750
},
{
"epoch": 20.32,
"learning_rate": 7.05448275862069e-06,
"loss": 0.0011,
"step": 4775
},
{
"epoch": 20.43,
"learning_rate": 7.037241379310346e-06,
"loss": 0.0011,
"step": 4800
},
{
"epoch": 20.53,
"learning_rate": 7.0200000000000006e-06,
"loss": 0.001,
"step": 4825
},
{
"epoch": 20.64,
"learning_rate": 7.002758620689655e-06,
"loss": 0.0011,
"step": 4850
},
{
"epoch": 20.74,
"learning_rate": 6.98551724137931e-06,
"loss": 0.0013,
"step": 4875
},
{
"epoch": 20.85,
"learning_rate": 6.968275862068966e-06,
"loss": 0.0018,
"step": 4900
},
{
"epoch": 20.96,
"learning_rate": 6.951034482758622e-06,
"loss": 0.0015,
"step": 4925
},
{
"epoch": 21.06,
"learning_rate": 6.933793103448277e-06,
"loss": 0.0011,
"step": 4950
},
{
"epoch": 21.17,
"learning_rate": 6.916551724137932e-06,
"loss": 0.0019,
"step": 4975
},
{
"epoch": 21.28,
"learning_rate": 6.899310344827586e-06,
"loss": 0.0017,
"step": 5000
},
{
"epoch": 21.28,
"eval_cer": 12.110361172651862,
"eval_loss": 0.5182740688323975,
"eval_runtime": 2246.2589,
"eval_samples_per_second": 0.84,
"eval_steps_per_second": 0.053,
"eval_wer": 35.27965916539218,
"step": 5000
},
{
"epoch": 21.38,
"learning_rate": 6.882068965517242e-06,
"loss": 0.0016,
"step": 5025
},
{
"epoch": 21.49,
"learning_rate": 6.864827586206897e-06,
"loss": 0.0011,
"step": 5050
},
{
"epoch": 21.6,
"learning_rate": 6.847586206896553e-06,
"loss": 0.0014,
"step": 5075
},
{
"epoch": 21.7,
"learning_rate": 6.8303448275862075e-06,
"loss": 0.0016,
"step": 5100
},
{
"epoch": 21.81,
"learning_rate": 6.813103448275863e-06,
"loss": 0.002,
"step": 5125
},
{
"epoch": 21.91,
"learning_rate": 6.795862068965517e-06,
"loss": 0.0012,
"step": 5150
},
{
"epoch": 22.02,
"learning_rate": 6.778620689655173e-06,
"loss": 0.0018,
"step": 5175
},
{
"epoch": 22.13,
"learning_rate": 6.761379310344828e-06,
"loss": 0.0017,
"step": 5200
},
{
"epoch": 22.23,
"learning_rate": 6.7441379310344836e-06,
"loss": 0.0026,
"step": 5225
},
{
"epoch": 22.34,
"learning_rate": 6.7268965517241384e-06,
"loss": 0.002,
"step": 5250
},
{
"epoch": 22.45,
"learning_rate": 6.709655172413793e-06,
"loss": 0.0031,
"step": 5275
},
{
"epoch": 22.55,
"learning_rate": 6.692413793103448e-06,
"loss": 0.0017,
"step": 5300
},
{
"epoch": 22.66,
"learning_rate": 6.675172413793104e-06,
"loss": 0.0023,
"step": 5325
},
{
"epoch": 22.77,
"learning_rate": 6.65793103448276e-06,
"loss": 0.0016,
"step": 5350
},
{
"epoch": 22.87,
"learning_rate": 6.6406896551724145e-06,
"loss": 0.0023,
"step": 5375
},
{
"epoch": 22.98,
"learning_rate": 6.62344827586207e-06,
"loss": 0.0029,
"step": 5400
},
{
"epoch": 23.09,
"learning_rate": 6.606206896551724e-06,
"loss": 0.0021,
"step": 5425
},
{
"epoch": 23.19,
"learning_rate": 6.58896551724138e-06,
"loss": 0.0022,
"step": 5450
},
{
"epoch": 23.3,
"learning_rate": 6.571724137931035e-06,
"loss": 0.0023,
"step": 5475
},
{
"epoch": 23.4,
"learning_rate": 6.5544827586206905e-06,
"loss": 0.0017,
"step": 5500
},
{
"epoch": 23.51,
"learning_rate": 6.537241379310345e-06,
"loss": 0.002,
"step": 5525
},
{
"epoch": 23.62,
"learning_rate": 6.520000000000001e-06,
"loss": 0.0021,
"step": 5550
},
{
"epoch": 23.72,
"learning_rate": 6.502758620689655e-06,
"loss": 0.0026,
"step": 5575
},
{
"epoch": 23.83,
"learning_rate": 6.485517241379311e-06,
"loss": 0.0028,
"step": 5600
},
{
"epoch": 23.94,
"learning_rate": 6.468275862068966e-06,
"loss": 0.0036,
"step": 5625
},
{
"epoch": 24.04,
"learning_rate": 6.4510344827586214e-06,
"loss": 0.0028,
"step": 5650
},
{
"epoch": 24.15,
"learning_rate": 6.433793103448276e-06,
"loss": 0.0029,
"step": 5675
},
{
"epoch": 24.26,
"learning_rate": 6.416551724137931e-06,
"loss": 0.0019,
"step": 5700
},
{
"epoch": 24.36,
"learning_rate": 6.399310344827587e-06,
"loss": 0.002,
"step": 5725
},
{
"epoch": 24.47,
"learning_rate": 6.382068965517242e-06,
"loss": 0.0021,
"step": 5750
},
{
"epoch": 24.57,
"learning_rate": 6.3648275862068975e-06,
"loss": 0.002,
"step": 5775
},
{
"epoch": 24.68,
"learning_rate": 6.347586206896552e-06,
"loss": 0.0018,
"step": 5800
},
{
"epoch": 24.79,
"learning_rate": 6.330344827586208e-06,
"loss": 0.0015,
"step": 5825
},
{
"epoch": 24.89,
"learning_rate": 6.313103448275862e-06,
"loss": 0.0021,
"step": 5850
},
{
"epoch": 25.0,
"learning_rate": 6.295862068965518e-06,
"loss": 0.0016,
"step": 5875
},
{
"epoch": 25.11,
"learning_rate": 6.278620689655173e-06,
"loss": 0.0017,
"step": 5900
},
{
"epoch": 25.21,
"learning_rate": 6.261379310344828e-06,
"loss": 0.0022,
"step": 5925
},
{
"epoch": 25.32,
"learning_rate": 6.244137931034483e-06,
"loss": 0.0012,
"step": 5950
},
{
"epoch": 25.43,
"learning_rate": 6.226896551724139e-06,
"loss": 0.0011,
"step": 5975
},
{
"epoch": 25.53,
"learning_rate": 6.209655172413793e-06,
"loss": 0.0016,
"step": 6000
},
{
"epoch": 25.53,
"eval_cer": 11.745410204887767,
"eval_loss": 0.5422167778015137,
"eval_runtime": 2139.2131,
"eval_samples_per_second": 0.882,
"eval_steps_per_second": 0.055,
"eval_wer": 35.43259777146602,
"step": 6000
},
{
"epoch": 25.64,
"learning_rate": 6.192413793103449e-06,
"loss": 0.0015,
"step": 6025
},
{
"epoch": 25.74,
"learning_rate": 6.175172413793104e-06,
"loss": 0.0019,
"step": 6050
},
{
"epoch": 25.85,
"learning_rate": 6.157931034482759e-06,
"loss": 0.0017,
"step": 6075
},
{
"epoch": 25.96,
"learning_rate": 6.140689655172414e-06,
"loss": 0.0015,
"step": 6100
},
{
"epoch": 26.06,
"learning_rate": 6.123448275862069e-06,
"loss": 0.0015,
"step": 6125
},
{
"epoch": 26.17,
"learning_rate": 6.106206896551725e-06,
"loss": 0.0016,
"step": 6150
},
{
"epoch": 26.28,
"learning_rate": 6.08896551724138e-06,
"loss": 0.0011,
"step": 6175
},
{
"epoch": 26.38,
"learning_rate": 6.071724137931035e-06,
"loss": 0.0015,
"step": 6200
},
{
"epoch": 26.49,
"learning_rate": 6.05448275862069e-06,
"loss": 0.0017,
"step": 6225
},
{
"epoch": 26.6,
"learning_rate": 6.037241379310346e-06,
"loss": 0.0013,
"step": 6250
},
{
"epoch": 26.7,
"learning_rate": 6.02e-06,
"loss": 0.0021,
"step": 6275
},
{
"epoch": 26.81,
"learning_rate": 6.002758620689656e-06,
"loss": 0.0017,
"step": 6300
},
{
"epoch": 26.91,
"learning_rate": 5.9855172413793105e-06,
"loss": 0.0017,
"step": 6325
},
{
"epoch": 27.02,
"learning_rate": 5.968275862068966e-06,
"loss": 0.0021,
"step": 6350
},
{
"epoch": 27.13,
"learning_rate": 5.951034482758621e-06,
"loss": 0.0008,
"step": 6375
},
{
"epoch": 27.23,
"learning_rate": 5.933793103448277e-06,
"loss": 0.0008,
"step": 6400
},
{
"epoch": 27.34,
"learning_rate": 5.916551724137931e-06,
"loss": 0.0007,
"step": 6425
},
{
"epoch": 27.45,
"learning_rate": 5.899310344827587e-06,
"loss": 0.0008,
"step": 6450
},
{
"epoch": 27.55,
"learning_rate": 5.8820689655172415e-06,
"loss": 0.0011,
"step": 6475
},
{
"epoch": 27.66,
"learning_rate": 5.864827586206897e-06,
"loss": 0.0016,
"step": 6500
},
{
"epoch": 27.77,
"learning_rate": 5.847586206896552e-06,
"loss": 0.0011,
"step": 6525
},
{
"epoch": 27.87,
"learning_rate": 5.830344827586208e-06,
"loss": 0.0013,
"step": 6550
},
{
"epoch": 27.98,
"learning_rate": 5.813103448275863e-06,
"loss": 0.0018,
"step": 6575
},
{
"epoch": 28.09,
"learning_rate": 5.7958620689655175e-06,
"loss": 0.002,
"step": 6600
},
{
"epoch": 28.19,
"learning_rate": 5.778620689655173e-06,
"loss": 0.0012,
"step": 6625
},
{
"epoch": 28.3,
"learning_rate": 5.761379310344828e-06,
"loss": 0.001,
"step": 6650
},
{
"epoch": 28.4,
"learning_rate": 5.744137931034484e-06,
"loss": 0.0012,
"step": 6675
},
{
"epoch": 28.51,
"learning_rate": 5.726896551724138e-06,
"loss": 0.0012,
"step": 6700
},
{
"epoch": 28.62,
"learning_rate": 5.7096551724137936e-06,
"loss": 0.0012,
"step": 6725
},
{
"epoch": 28.72,
"learning_rate": 5.692413793103448e-06,
"loss": 0.0018,
"step": 6750
},
{
"epoch": 28.83,
"learning_rate": 5.675172413793104e-06,
"loss": 0.0016,
"step": 6775
},
{
"epoch": 28.94,
"learning_rate": 5.657931034482759e-06,
"loss": 0.0011,
"step": 6800
},
{
"epoch": 29.04,
"learning_rate": 5.640689655172415e-06,
"loss": 0.0012,
"step": 6825
},
{
"epoch": 29.15,
"learning_rate": 5.623448275862069e-06,
"loss": 0.0012,
"step": 6850
},
{
"epoch": 29.26,
"learning_rate": 5.6062068965517245e-06,
"loss": 0.0009,
"step": 6875
},
{
"epoch": 29.36,
"learning_rate": 5.588965517241379e-06,
"loss": 0.0012,
"step": 6900
},
{
"epoch": 29.47,
"learning_rate": 5.571724137931035e-06,
"loss": 0.0011,
"step": 6925
},
{
"epoch": 29.57,
"learning_rate": 5.554482758620691e-06,
"loss": 0.0013,
"step": 6950
},
{
"epoch": 29.68,
"learning_rate": 5.537241379310346e-06,
"loss": 0.0011,
"step": 6975
},
{
"epoch": 29.79,
"learning_rate": 5.5200000000000005e-06,
"loss": 0.0011,
"step": 7000
},
{
"epoch": 29.79,
"eval_cer": 11.51956995895375,
"eval_loss": 0.5360536575317383,
"eval_runtime": 2155.9894,
"eval_samples_per_second": 0.875,
"eval_steps_per_second": 0.055,
"eval_wer": 34.531352414245134,
"step": 7000
},
{
"epoch": 29.89,
"learning_rate": 5.502758620689655e-06,
"loss": 0.0009,
"step": 7025
},
{
"epoch": 30.0,
"learning_rate": 5.485517241379311e-06,
"loss": 0.0019,
"step": 7050
},
{
"epoch": 30.11,
"learning_rate": 5.468275862068966e-06,
"loss": 0.0011,
"step": 7075
},
{
"epoch": 30.21,
"learning_rate": 5.451034482758622e-06,
"loss": 0.0011,
"step": 7100
},
{
"epoch": 30.32,
"learning_rate": 5.433793103448276e-06,
"loss": 0.0014,
"step": 7125
},
{
"epoch": 30.43,
"learning_rate": 5.4165517241379314e-06,
"loss": 0.0013,
"step": 7150
},
{
"epoch": 30.53,
"learning_rate": 5.399310344827586e-06,
"loss": 0.001,
"step": 7175
},
{
"epoch": 30.64,
"learning_rate": 5.382068965517242e-06,
"loss": 0.0011,
"step": 7200
},
{
"epoch": 30.74,
"learning_rate": 5.364827586206897e-06,
"loss": 0.001,
"step": 7225
},
{
"epoch": 30.85,
"learning_rate": 5.347586206896553e-06,
"loss": 0.0014,
"step": 7250
},
{
"epoch": 30.96,
"learning_rate": 5.330344827586207e-06,
"loss": 0.0022,
"step": 7275
},
{
"epoch": 31.06,
"learning_rate": 5.313103448275862e-06,
"loss": 0.0012,
"step": 7300
},
{
"epoch": 31.17,
"learning_rate": 5.295862068965517e-06,
"loss": 0.001,
"step": 7325
},
{
"epoch": 31.28,
"learning_rate": 5.278620689655173e-06,
"loss": 0.001,
"step": 7350
},
{
"epoch": 31.38,
"learning_rate": 5.261379310344829e-06,
"loss": 0.0017,
"step": 7375
},
{
"epoch": 31.49,
"learning_rate": 5.2441379310344835e-06,
"loss": 0.0016,
"step": 7400
},
{
"epoch": 31.6,
"learning_rate": 5.226896551724138e-06,
"loss": 0.0012,
"step": 7425
},
{
"epoch": 31.7,
"learning_rate": 5.209655172413793e-06,
"loss": 0.0009,
"step": 7450
},
{
"epoch": 31.81,
"learning_rate": 5.192413793103449e-06,
"loss": 0.001,
"step": 7475
},
{
"epoch": 31.91,
"learning_rate": 5.175172413793104e-06,
"loss": 0.0011,
"step": 7500
},
{
"epoch": 32.02,
"learning_rate": 5.1579310344827596e-06,
"loss": 0.0009,
"step": 7525
},
{
"epoch": 32.13,
"learning_rate": 5.140689655172414e-06,
"loss": 0.0003,
"step": 7550
},
{
"epoch": 32.23,
"learning_rate": 5.123448275862069e-06,
"loss": 0.0007,
"step": 7575
},
{
"epoch": 32.34,
"learning_rate": 5.106206896551724e-06,
"loss": 0.0007,
"step": 7600
},
{
"epoch": 32.45,
"learning_rate": 5.08896551724138e-06,
"loss": 0.001,
"step": 7625
},
{
"epoch": 32.55,
"learning_rate": 5.071724137931035e-06,
"loss": 0.0009,
"step": 7650
},
{
"epoch": 32.66,
"learning_rate": 5.0544827586206905e-06,
"loss": 0.0007,
"step": 7675
},
{
"epoch": 32.77,
"learning_rate": 5.0372413793103445e-06,
"loss": 0.0007,
"step": 7700
},
{
"epoch": 32.87,
"learning_rate": 5.02e-06,
"loss": 0.0007,
"step": 7725
},
{
"epoch": 32.98,
"learning_rate": 5.002758620689656e-06,
"loss": 0.0007,
"step": 7750
},
{
"epoch": 33.09,
"learning_rate": 4.985517241379311e-06,
"loss": 0.0009,
"step": 7775
},
{
"epoch": 33.19,
"learning_rate": 4.968275862068966e-06,
"loss": 0.0012,
"step": 7800
},
{
"epoch": 33.3,
"learning_rate": 4.951034482758621e-06,
"loss": 0.0009,
"step": 7825
},
{
"epoch": 33.4,
"learning_rate": 4.933793103448276e-06,
"loss": 0.0004,
"step": 7850
},
{
"epoch": 33.51,
"learning_rate": 4.916551724137931e-06,
"loss": 0.0004,
"step": 7875
},
{
"epoch": 33.62,
"learning_rate": 4.899310344827586e-06,
"loss": 0.0002,
"step": 7900
},
{
"epoch": 33.72,
"learning_rate": 4.882068965517242e-06,
"loss": 0.0008,
"step": 7925
},
{
"epoch": 33.83,
"learning_rate": 4.8648275862068974e-06,
"loss": 0.0003,
"step": 7950
},
{
"epoch": 33.94,
"learning_rate": 4.847586206896552e-06,
"loss": 0.0002,
"step": 7975
},
{
"epoch": 34.04,
"learning_rate": 4.830344827586207e-06,
"loss": 0.0004,
"step": 8000
},
{
"epoch": 34.04,
"eval_cer": 11.365002490253662,
"eval_loss": 0.5406305193901062,
"eval_runtime": 2161.7975,
"eval_samples_per_second": 0.872,
"eval_steps_per_second": 0.055,
"eval_wer": 34.09984706139393,
"step": 8000
},
{
"epoch": 34.15,
"learning_rate": 4.813103448275863e-06,
"loss": 0.0001,
"step": 8025
},
{
"epoch": 34.26,
"learning_rate": 4.795862068965518e-06,
"loss": 0.0005,
"step": 8050
},
{
"epoch": 34.36,
"learning_rate": 4.778620689655173e-06,
"loss": 0.0002,
"step": 8075
},
{
"epoch": 34.47,
"learning_rate": 4.761379310344828e-06,
"loss": 0.0004,
"step": 8100
},
{
"epoch": 34.57,
"learning_rate": 4.744137931034483e-06,
"loss": 0.001,
"step": 8125
},
{
"epoch": 34.68,
"learning_rate": 4.726896551724138e-06,
"loss": 0.0003,
"step": 8150
},
{
"epoch": 34.79,
"learning_rate": 4.709655172413794e-06,
"loss": 0.0003,
"step": 8175
},
{
"epoch": 34.89,
"learning_rate": 4.692413793103449e-06,
"loss": 0.0003,
"step": 8200
},
{
"epoch": 35.0,
"learning_rate": 4.6751724137931035e-06,
"loss": 0.0004,
"step": 8225
},
{
"epoch": 35.11,
"learning_rate": 4.657931034482759e-06,
"loss": 0.0001,
"step": 8250
},
{
"epoch": 35.21,
"learning_rate": 4.640689655172414e-06,
"loss": 0.0001,
"step": 8275
},
{
"epoch": 35.32,
"learning_rate": 4.623448275862069e-06,
"loss": 0.0003,
"step": 8300
},
{
"epoch": 35.43,
"learning_rate": 4.606206896551724e-06,
"loss": 0.0002,
"step": 8325
},
{
"epoch": 35.53,
"learning_rate": 4.58896551724138e-06,
"loss": 0.0001,
"step": 8350
},
{
"epoch": 35.64,
"learning_rate": 4.571724137931035e-06,
"loss": 0.0005,
"step": 8375
},
{
"epoch": 35.74,
"learning_rate": 4.55448275862069e-06,
"loss": 0.0003,
"step": 8400
},
{
"epoch": 35.85,
"learning_rate": 4.537241379310345e-06,
"loss": 0.001,
"step": 8425
},
{
"epoch": 35.96,
"learning_rate": 4.520000000000001e-06,
"loss": 0.0004,
"step": 8450
},
{
"epoch": 36.06,
"learning_rate": 4.502758620689656e-06,
"loss": 0.0004,
"step": 8475
},
{
"epoch": 36.17,
"learning_rate": 4.4855172413793105e-06,
"loss": 0.0006,
"step": 8500
},
{
"epoch": 36.28,
"learning_rate": 4.468275862068966e-06,
"loss": 0.0002,
"step": 8525
},
{
"epoch": 36.38,
"learning_rate": 4.451034482758621e-06,
"loss": 0.0003,
"step": 8550
},
{
"epoch": 36.49,
"learning_rate": 4.433793103448276e-06,
"loss": 0.0003,
"step": 8575
},
{
"epoch": 36.6,
"learning_rate": 4.416551724137932e-06,
"loss": 0.0004,
"step": 8600
},
{
"epoch": 36.7,
"learning_rate": 4.3993103448275866e-06,
"loss": 0.0003,
"step": 8625
},
{
"epoch": 36.81,
"learning_rate": 4.3820689655172414e-06,
"loss": 0.0008,
"step": 8650
},
{
"epoch": 36.91,
"learning_rate": 4.364827586206897e-06,
"loss": 0.0006,
"step": 8675
},
{
"epoch": 37.02,
"learning_rate": 4.347586206896552e-06,
"loss": 0.0005,
"step": 8700
},
{
"epoch": 37.13,
"learning_rate": 4.330344827586207e-06,
"loss": 0.001,
"step": 8725
},
{
"epoch": 37.23,
"learning_rate": 4.313103448275863e-06,
"loss": 0.0004,
"step": 8750
},
{
"epoch": 37.34,
"learning_rate": 4.2958620689655175e-06,
"loss": 0.0004,
"step": 8775
},
{
"epoch": 37.45,
"learning_rate": 4.278620689655173e-06,
"loss": 0.0005,
"step": 8800
},
{
"epoch": 37.55,
"learning_rate": 4.261379310344828e-06,
"loss": 0.0005,
"step": 8825
},
{
"epoch": 37.66,
"learning_rate": 4.244137931034483e-06,
"loss": 0.0003,
"step": 8850
},
{
"epoch": 37.77,
"learning_rate": 4.226896551724139e-06,
"loss": 0.0009,
"step": 8875
},
{
"epoch": 37.87,
"learning_rate": 4.2096551724137935e-06,
"loss": 0.0004,
"step": 8900
},
{
"epoch": 37.98,
"learning_rate": 4.192413793103448e-06,
"loss": 0.0006,
"step": 8925
},
{
"epoch": 38.09,
"learning_rate": 4.175172413793104e-06,
"loss": 0.0006,
"step": 8950
},
{
"epoch": 38.19,
"learning_rate": 4.157931034482759e-06,
"loss": 0.0008,
"step": 8975
},
{
"epoch": 38.3,
"learning_rate": 4.140689655172414e-06,
"loss": 0.0006,
"step": 9000
},
{
"epoch": 38.3,
"eval_cer": 11.291153588541398,
"eval_loss": 0.5540400147438049,
"eval_runtime": 2151.4681,
"eval_samples_per_second": 0.877,
"eval_steps_per_second": 0.055,
"eval_wer": 33.86497705920909,
"step": 9000
},
{
"epoch": 38.4,
"learning_rate": 4.1234482758620696e-06,
"loss": 0.0005,
"step": 9025
},
{
"epoch": 38.51,
"learning_rate": 4.1062068965517244e-06,
"loss": 0.0007,
"step": 9050
},
{
"epoch": 38.62,
"learning_rate": 4.088965517241379e-06,
"loss": 0.0004,
"step": 9075
},
{
"epoch": 38.72,
"learning_rate": 4.071724137931035e-06,
"loss": 0.0004,
"step": 9100
},
{
"epoch": 38.83,
"learning_rate": 4.05448275862069e-06,
"loss": 0.0004,
"step": 9125
},
{
"epoch": 38.94,
"learning_rate": 4.037241379310346e-06,
"loss": 0.0002,
"step": 9150
},
{
"epoch": 39.04,
"learning_rate": 4.0200000000000005e-06,
"loss": 0.0005,
"step": 9175
},
{
"epoch": 39.15,
"learning_rate": 4.002758620689655e-06,
"loss": 0.0001,
"step": 9200
},
{
"epoch": 39.26,
"learning_rate": 3.985517241379311e-06,
"loss": 0.0001,
"step": 9225
},
{
"epoch": 39.36,
"learning_rate": 3.968275862068966e-06,
"loss": 0.0003,
"step": 9250
},
{
"epoch": 39.47,
"learning_rate": 3.951034482758621e-06,
"loss": 0.0003,
"step": 9275
},
{
"epoch": 39.57,
"learning_rate": 3.9337931034482765e-06,
"loss": 0.0005,
"step": 9300
},
{
"epoch": 39.68,
"learning_rate": 3.916551724137931e-06,
"loss": 0.0002,
"step": 9325
},
{
"epoch": 39.79,
"learning_rate": 3.899310344827586e-06,
"loss": 0.0005,
"step": 9350
},
{
"epoch": 39.89,
"learning_rate": 3.882068965517242e-06,
"loss": 0.0009,
"step": 9375
},
{
"epoch": 40.0,
"learning_rate": 3.864827586206897e-06,
"loss": 0.0007,
"step": 9400
},
{
"epoch": 40.11,
"learning_rate": 3.847586206896552e-06,
"loss": 0.0008,
"step": 9425
},
{
"epoch": 40.21,
"learning_rate": 3.8303448275862074e-06,
"loss": 0.0003,
"step": 9450
},
{
"epoch": 40.32,
"learning_rate": 3.8131034482758623e-06,
"loss": 0.0009,
"step": 9475
},
{
"epoch": 40.43,
"learning_rate": 3.795862068965517e-06,
"loss": 0.001,
"step": 9500
},
{
"epoch": 40.53,
"learning_rate": 3.7786206896551725e-06,
"loss": 0.0006,
"step": 9525
},
{
"epoch": 40.64,
"learning_rate": 3.7613793103448278e-06,
"loss": 0.0005,
"step": 9550
},
{
"epoch": 40.74,
"learning_rate": 3.7441379310344835e-06,
"loss": 0.0003,
"step": 9575
},
{
"epoch": 40.85,
"learning_rate": 3.7268965517241383e-06,
"loss": 0.001,
"step": 9600
},
{
"epoch": 40.96,
"learning_rate": 3.7096551724137936e-06,
"loss": 0.0005,
"step": 9625
},
{
"epoch": 41.06,
"learning_rate": 3.6924137931034485e-06,
"loss": 0.0006,
"step": 9650
},
{
"epoch": 41.17,
"learning_rate": 3.675172413793104e-06,
"loss": 0.0005,
"step": 9675
},
{
"epoch": 41.28,
"learning_rate": 3.657931034482759e-06,
"loss": 0.001,
"step": 9700
},
{
"epoch": 41.38,
"learning_rate": 3.640689655172414e-06,
"loss": 0.0006,
"step": 9725
},
{
"epoch": 41.49,
"learning_rate": 3.6234482758620693e-06,
"loss": 0.0003,
"step": 9750
},
{
"epoch": 41.6,
"learning_rate": 3.6062068965517246e-06,
"loss": 0.0003,
"step": 9775
},
{
"epoch": 41.7,
"learning_rate": 3.5889655172413794e-06,
"loss": 0.0002,
"step": 9800
},
{
"epoch": 41.81,
"learning_rate": 3.5717241379310347e-06,
"loss": 0.0002,
"step": 9825
},
{
"epoch": 41.91,
"learning_rate": 3.55448275862069e-06,
"loss": 0.0003,
"step": 9850
},
{
"epoch": 42.02,
"learning_rate": 3.537241379310345e-06,
"loss": 0.0005,
"step": 9875
},
{
"epoch": 42.13,
"learning_rate": 3.52e-06,
"loss": 0.0004,
"step": 9900
},
{
"epoch": 42.23,
"learning_rate": 3.502758620689655e-06,
"loss": 0.0001,
"step": 9925
},
{
"epoch": 42.34,
"learning_rate": 3.4855172413793103e-06,
"loss": 0.0004,
"step": 9950
},
{
"epoch": 42.45,
"learning_rate": 3.468275862068966e-06,
"loss": 0.0002,
"step": 9975
},
{
"epoch": 42.55,
"learning_rate": 3.4510344827586214e-06,
"loss": 0.0002,
"step": 10000
},
{
"epoch": 42.55,
"eval_cer": 11.533309289504869,
"eval_loss": 0.5747844576835632,
"eval_runtime": 2192.2081,
"eval_samples_per_second": 0.86,
"eval_steps_per_second": 0.054,
"eval_wer": 34.08892287524579,
"step": 10000
},
{
"epoch": 42.66,
"learning_rate": 3.4337931034482762e-06,
"loss": 0.0002,
"step": 10025
},
{
"epoch": 42.77,
"learning_rate": 3.4165517241379315e-06,
"loss": 0.0005,
"step": 10050
},
{
"epoch": 42.87,
"learning_rate": 3.3993103448275864e-06,
"loss": 0.0006,
"step": 10075
},
{
"epoch": 42.98,
"learning_rate": 3.3820689655172417e-06,
"loss": 0.0007,
"step": 10100
},
{
"epoch": 43.09,
"learning_rate": 3.364827586206897e-06,
"loss": 0.0002,
"step": 10125
},
{
"epoch": 43.19,
"learning_rate": 3.347586206896552e-06,
"loss": 0.0009,
"step": 10150
},
{
"epoch": 43.3,
"learning_rate": 3.330344827586207e-06,
"loss": 0.0005,
"step": 10175
},
{
"epoch": 43.4,
"learning_rate": 3.3131034482758624e-06,
"loss": 0.0005,
"step": 10200
},
{
"epoch": 43.51,
"learning_rate": 3.2958620689655173e-06,
"loss": 0.0005,
"step": 10225
},
{
"epoch": 43.62,
"learning_rate": 3.2786206896551726e-06,
"loss": 0.0002,
"step": 10250
},
{
"epoch": 43.72,
"learning_rate": 3.261379310344828e-06,
"loss": 0.0003,
"step": 10275
},
{
"epoch": 43.83,
"learning_rate": 3.2441379310344828e-06,
"loss": 0.0002,
"step": 10300
},
{
"epoch": 43.94,
"learning_rate": 3.226896551724138e-06,
"loss": 0.0001,
"step": 10325
},
{
"epoch": 44.04,
"learning_rate": 3.209655172413793e-06,
"loss": 0.0001,
"step": 10350
},
{
"epoch": 44.15,
"learning_rate": 3.1924137931034486e-06,
"loss": 0.0002,
"step": 10375
},
{
"epoch": 44.26,
"learning_rate": 3.175172413793104e-06,
"loss": 0.0002,
"step": 10400
},
{
"epoch": 44.36,
"learning_rate": 3.1579310344827592e-06,
"loss": 0.0002,
"step": 10425
},
{
"epoch": 44.47,
"learning_rate": 3.140689655172414e-06,
"loss": 0.0006,
"step": 10450
},
{
"epoch": 44.57,
"learning_rate": 3.1234482758620694e-06,
"loss": 0.0001,
"step": 10475
},
{
"epoch": 44.68,
"learning_rate": 3.1062068965517243e-06,
"loss": 0.0003,
"step": 10500
},
{
"epoch": 44.79,
"learning_rate": 3.0889655172413796e-06,
"loss": 0.0001,
"step": 10525
},
{
"epoch": 44.89,
"learning_rate": 3.071724137931035e-06,
"loss": 0.0001,
"step": 10550
},
{
"epoch": 45.0,
"learning_rate": 3.0544827586206897e-06,
"loss": 0.0001,
"step": 10575
},
{
"epoch": 45.11,
"learning_rate": 3.037241379310345e-06,
"loss": 0.0002,
"step": 10600
},
{
"epoch": 45.21,
"learning_rate": 3.0200000000000003e-06,
"loss": 0.0009,
"step": 10625
},
{
"epoch": 45.32,
"learning_rate": 3.002758620689655e-06,
"loss": 0.0002,
"step": 10650
},
{
"epoch": 45.43,
"learning_rate": 2.9855172413793105e-06,
"loss": 0.0002,
"step": 10675
},
{
"epoch": 45.53,
"learning_rate": 2.9682758620689658e-06,
"loss": 0.0005,
"step": 10700
},
{
"epoch": 45.64,
"learning_rate": 2.9510344827586206e-06,
"loss": 0.0003,
"step": 10725
},
{
"epoch": 45.74,
"learning_rate": 2.933793103448276e-06,
"loss": 0.0005,
"step": 10750
},
{
"epoch": 45.85,
"learning_rate": 2.9165517241379316e-06,
"loss": 0.0006,
"step": 10775
},
{
"epoch": 45.96,
"learning_rate": 2.8993103448275865e-06,
"loss": 0.0002,
"step": 10800
},
{
"epoch": 46.06,
"learning_rate": 2.882068965517242e-06,
"loss": 0.0002,
"step": 10825
},
{
"epoch": 46.17,
"learning_rate": 2.864827586206897e-06,
"loss": 0.0001,
"step": 10850
},
{
"epoch": 46.28,
"learning_rate": 2.847586206896552e-06,
"loss": 0.0002,
"step": 10875
},
{
"epoch": 46.38,
"learning_rate": 2.8303448275862073e-06,
"loss": 0.0001,
"step": 10900
},
{
"epoch": 46.49,
"learning_rate": 2.813103448275862e-06,
"loss": 0.0007,
"step": 10925
},
{
"epoch": 46.6,
"learning_rate": 2.7958620689655174e-06,
"loss": 0.0005,
"step": 10950
},
{
"epoch": 46.7,
"learning_rate": 2.7786206896551727e-06,
"loss": 0.0006,
"step": 10975
},
{
"epoch": 46.81,
"learning_rate": 2.7613793103448276e-06,
"loss": 0.0003,
"step": 11000
},
{
"epoch": 46.81,
"eval_cer": 11.489515173373178,
"eval_loss": 0.5770519971847534,
"eval_runtime": 2190.3207,
"eval_samples_per_second": 0.861,
"eval_steps_per_second": 0.054,
"eval_wer": 34.564124972689534,
"step": 11000
},
{
"epoch": 46.91,
"learning_rate": 2.744137931034483e-06,
"loss": 0.0002,
"step": 11025
},
{
"epoch": 47.02,
"learning_rate": 2.726896551724138e-06,
"loss": 0.0003,
"step": 11050
},
{
"epoch": 47.13,
"learning_rate": 2.709655172413793e-06,
"loss": 0.0009,
"step": 11075
},
{
"epoch": 47.23,
"learning_rate": 2.6924137931034483e-06,
"loss": 0.0003,
"step": 11100
},
{
"epoch": 47.34,
"learning_rate": 2.6751724137931036e-06,
"loss": 0.0001,
"step": 11125
},
{
"epoch": 47.45,
"learning_rate": 2.6579310344827585e-06,
"loss": 0.0001,
"step": 11150
},
{
"epoch": 47.55,
"learning_rate": 2.6406896551724142e-06,
"loss": 0.0003,
"step": 11175
},
{
"epoch": 47.66,
"learning_rate": 2.6234482758620695e-06,
"loss": 0.0001,
"step": 11200
},
{
"epoch": 47.77,
"learning_rate": 2.6062068965517244e-06,
"loss": 0.0002,
"step": 11225
},
{
"epoch": 47.87,
"learning_rate": 2.5889655172413797e-06,
"loss": 0.0002,
"step": 11250
},
{
"epoch": 47.98,
"learning_rate": 2.571724137931035e-06,
"loss": 0.0001,
"step": 11275
},
{
"epoch": 48.09,
"learning_rate": 2.55448275862069e-06,
"loss": 0.0001,
"step": 11300
},
{
"epoch": 48.19,
"learning_rate": 2.537241379310345e-06,
"loss": 0.0001,
"step": 11325
},
{
"epoch": 48.3,
"learning_rate": 2.52e-06,
"loss": 0.0001,
"step": 11350
},
{
"epoch": 48.4,
"learning_rate": 2.5027586206896553e-06,
"loss": 0.0005,
"step": 11375
},
{
"epoch": 48.51,
"learning_rate": 2.4855172413793106e-06,
"loss": 0.0001,
"step": 11400
},
{
"epoch": 48.62,
"learning_rate": 2.4682758620689655e-06,
"loss": 0.0001,
"step": 11425
},
{
"epoch": 48.72,
"learning_rate": 2.4510344827586208e-06,
"loss": 0.0002,
"step": 11450
},
{
"epoch": 48.83,
"learning_rate": 2.433793103448276e-06,
"loss": 0.0001,
"step": 11475
},
{
"epoch": 48.94,
"learning_rate": 2.4165517241379314e-06,
"loss": 0.0003,
"step": 11500
},
{
"epoch": 49.04,
"learning_rate": 2.3993103448275866e-06,
"loss": 0.0001,
"step": 11525
},
{
"epoch": 49.15,
"learning_rate": 2.3820689655172415e-06,
"loss": 0.0001,
"step": 11550
},
{
"epoch": 49.26,
"learning_rate": 2.364827586206897e-06,
"loss": 0.0001,
"step": 11575
},
{
"epoch": 49.36,
"learning_rate": 2.3475862068965517e-06,
"loss": 0.0001,
"step": 11600
},
{
"epoch": 49.47,
"learning_rate": 2.330344827586207e-06,
"loss": 0.0,
"step": 11625
},
{
"epoch": 49.57,
"learning_rate": 2.3131034482758623e-06,
"loss": 0.0,
"step": 11650
},
{
"epoch": 49.68,
"learning_rate": 2.2958620689655176e-06,
"loss": 0.0,
"step": 11675
},
{
"epoch": 49.79,
"learning_rate": 2.278620689655173e-06,
"loss": 0.0002,
"step": 11700
},
{
"epoch": 49.89,
"learning_rate": 2.2613793103448277e-06,
"loss": 0.0001,
"step": 11725
},
{
"epoch": 50.0,
"learning_rate": 2.244137931034483e-06,
"loss": 0.0,
"step": 11750
},
{
"epoch": 50.11,
"learning_rate": 2.2268965517241383e-06,
"loss": 0.0,
"step": 11775
},
{
"epoch": 50.21,
"learning_rate": 2.209655172413793e-06,
"loss": 0.0,
"step": 11800
},
{
"epoch": 50.32,
"learning_rate": 2.1924137931034485e-06,
"loss": 0.0,
"step": 11825
},
{
"epoch": 50.43,
"learning_rate": 2.1751724137931033e-06,
"loss": 0.0,
"step": 11850
},
{
"epoch": 50.53,
"learning_rate": 2.157931034482759e-06,
"loss": 0.0,
"step": 11875
},
{
"epoch": 50.64,
"learning_rate": 2.140689655172414e-06,
"loss": 0.0,
"step": 11900
},
{
"epoch": 50.74,
"learning_rate": 2.1234482758620692e-06,
"loss": 0.0,
"step": 11925
},
{
"epoch": 50.85,
"learning_rate": 2.1062068965517245e-06,
"loss": 0.0,
"step": 11950
},
{
"epoch": 50.96,
"learning_rate": 2.0889655172413794e-06,
"loss": 0.0,
"step": 11975
},
{
"epoch": 51.06,
"learning_rate": 2.0717241379310347e-06,
"loss": 0.0,
"step": 12000
},
{
"epoch": 51.06,
"eval_cer": 11.207000188915794,
"eval_loss": 0.5809110999107361,
"eval_runtime": 2152.4674,
"eval_samples_per_second": 0.876,
"eval_steps_per_second": 0.055,
"eval_wer": 33.43347170635788,
"step": 12000
},
{
"epoch": 51.17,
"learning_rate": 2.0544827586206896e-06,
"loss": 0.0,
"step": 12025
},
{
"epoch": 51.28,
"learning_rate": 2.037241379310345e-06,
"loss": 0.0,
"step": 12050
},
{
"epoch": 51.38,
"learning_rate": 2.02e-06,
"loss": 0.0,
"step": 12075
},
{
"epoch": 51.49,
"learning_rate": 2.0027586206896554e-06,
"loss": 0.0,
"step": 12100
},
{
"epoch": 51.6,
"learning_rate": 1.9855172413793107e-06,
"loss": 0.0,
"step": 12125
},
{
"epoch": 51.7,
"learning_rate": 1.9682758620689656e-06,
"loss": 0.0,
"step": 12150
},
{
"epoch": 51.81,
"learning_rate": 1.951034482758621e-06,
"loss": 0.0,
"step": 12175
},
{
"epoch": 51.91,
"learning_rate": 1.933793103448276e-06,
"loss": 0.0,
"step": 12200
},
{
"epoch": 52.02,
"learning_rate": 1.916551724137931e-06,
"loss": 0.0,
"step": 12225
},
{
"epoch": 52.13,
"learning_rate": 1.8993103448275864e-06,
"loss": 0.0,
"step": 12250
},
{
"epoch": 52.23,
"learning_rate": 1.8820689655172416e-06,
"loss": 0.0,
"step": 12275
},
{
"epoch": 52.34,
"learning_rate": 1.8648275862068967e-06,
"loss": 0.0,
"step": 12300
},
{
"epoch": 52.45,
"learning_rate": 1.847586206896552e-06,
"loss": 0.0,
"step": 12325
},
{
"epoch": 52.55,
"learning_rate": 1.830344827586207e-06,
"loss": 0.0,
"step": 12350
},
{
"epoch": 52.66,
"learning_rate": 1.8131034482758622e-06,
"loss": 0.0,
"step": 12375
},
{
"epoch": 52.77,
"learning_rate": 1.7958620689655173e-06,
"loss": 0.0,
"step": 12400
},
{
"epoch": 52.87,
"learning_rate": 1.7786206896551726e-06,
"loss": 0.0,
"step": 12425
},
{
"epoch": 52.98,
"learning_rate": 1.7613793103448276e-06,
"loss": 0.0,
"step": 12450
},
{
"epoch": 53.09,
"learning_rate": 1.744137931034483e-06,
"loss": 0.0,
"step": 12475
},
{
"epoch": 53.19,
"learning_rate": 1.7268965517241382e-06,
"loss": 0.0,
"step": 12500
},
{
"epoch": 53.3,
"learning_rate": 1.7096551724137933e-06,
"loss": 0.0,
"step": 12525
},
{
"epoch": 53.4,
"learning_rate": 1.6924137931034484e-06,
"loss": 0.0,
"step": 12550
},
{
"epoch": 53.51,
"learning_rate": 1.6751724137931037e-06,
"loss": 0.0,
"step": 12575
},
{
"epoch": 53.62,
"learning_rate": 1.6579310344827588e-06,
"loss": 0.0,
"step": 12600
},
{
"epoch": 53.72,
"learning_rate": 1.6406896551724138e-06,
"loss": 0.0,
"step": 12625
},
{
"epoch": 53.83,
"learning_rate": 1.623448275862069e-06,
"loss": 0.0,
"step": 12650
},
{
"epoch": 53.94,
"learning_rate": 1.6062068965517244e-06,
"loss": 0.0,
"step": 12675
},
{
"epoch": 54.04,
"learning_rate": 1.5889655172413795e-06,
"loss": 0.0,
"step": 12700
},
{
"epoch": 54.15,
"learning_rate": 1.5717241379310346e-06,
"loss": 0.0,
"step": 12725
},
{
"epoch": 54.26,
"learning_rate": 1.55448275862069e-06,
"loss": 0.0,
"step": 12750
},
{
"epoch": 54.36,
"learning_rate": 1.537241379310345e-06,
"loss": 0.0,
"step": 12775
},
{
"epoch": 54.47,
"learning_rate": 1.52e-06,
"loss": 0.0,
"step": 12800
},
{
"epoch": 54.57,
"learning_rate": 1.5027586206896551e-06,
"loss": 0.0,
"step": 12825
},
{
"epoch": 54.68,
"learning_rate": 1.4855172413793104e-06,
"loss": 0.0,
"step": 12850
},
{
"epoch": 54.79,
"learning_rate": 1.4682758620689657e-06,
"loss": 0.0,
"step": 12875
},
{
"epoch": 54.89,
"learning_rate": 1.4510344827586208e-06,
"loss": 0.0,
"step": 12900
},
{
"epoch": 55.0,
"learning_rate": 1.433793103448276e-06,
"loss": 0.0,
"step": 12925
},
{
"epoch": 55.11,
"learning_rate": 1.4165517241379312e-06,
"loss": 0.0,
"step": 12950
},
{
"epoch": 55.21,
"learning_rate": 1.3993103448275863e-06,
"loss": 0.0,
"step": 12975
},
{
"epoch": 55.32,
"learning_rate": 1.3820689655172416e-06,
"loss": 0.0,
"step": 13000
},
{
"epoch": 55.32,
"eval_cer": 11.000910230649012,
"eval_loss": 0.5941275358200073,
"eval_runtime": 2139.688,
"eval_samples_per_second": 0.881,
"eval_steps_per_second": 0.055,
"eval_wer": 33.209525890321174,
"step": 13000
},
{
"epoch": 55.43,
"learning_rate": 1.3648275862068966e-06,
"loss": 0.0,
"step": 13025
},
{
"epoch": 55.53,
"learning_rate": 1.3475862068965517e-06,
"loss": 0.0,
"step": 13050
},
{
"epoch": 55.64,
"learning_rate": 1.3303448275862072e-06,
"loss": 0.0,
"step": 13075
},
{
"epoch": 55.74,
"learning_rate": 1.3131034482758623e-06,
"loss": 0.0,
"step": 13100
},
{
"epoch": 55.85,
"learning_rate": 1.2958620689655174e-06,
"loss": 0.0,
"step": 13125
},
{
"epoch": 55.96,
"learning_rate": 1.2786206896551725e-06,
"loss": 0.0,
"step": 13150
},
{
"epoch": 56.06,
"learning_rate": 1.2613793103448278e-06,
"loss": 0.0,
"step": 13175
},
{
"epoch": 56.17,
"learning_rate": 1.2441379310344829e-06,
"loss": 0.0,
"step": 13200
},
{
"epoch": 56.28,
"learning_rate": 1.226896551724138e-06,
"loss": 0.0,
"step": 13225
},
{
"epoch": 56.38,
"learning_rate": 1.2096551724137932e-06,
"loss": 0.0,
"step": 13250
},
{
"epoch": 56.49,
"learning_rate": 1.1924137931034483e-06,
"loss": 0.0,
"step": 13275
},
{
"epoch": 56.6,
"learning_rate": 1.1751724137931036e-06,
"loss": 0.0,
"step": 13300
},
{
"epoch": 56.7,
"learning_rate": 1.1579310344827587e-06,
"loss": 0.0,
"step": 13325
},
{
"epoch": 56.81,
"learning_rate": 1.140689655172414e-06,
"loss": 0.0,
"step": 13350
},
{
"epoch": 56.91,
"learning_rate": 1.123448275862069e-06,
"loss": 0.0,
"step": 13375
},
{
"epoch": 57.02,
"learning_rate": 1.1062068965517241e-06,
"loss": 0.0,
"step": 13400
},
{
"epoch": 57.13,
"learning_rate": 1.0889655172413794e-06,
"loss": 0.0,
"step": 13425
},
{
"epoch": 57.23,
"learning_rate": 1.0717241379310345e-06,
"loss": 0.0,
"step": 13450
},
{
"epoch": 57.34,
"learning_rate": 1.0544827586206898e-06,
"loss": 0.0,
"step": 13475
},
{
"epoch": 57.45,
"learning_rate": 1.0372413793103449e-06,
"loss": 0.0,
"step": 13500
},
{
"epoch": 57.55,
"learning_rate": 1.02e-06,
"loss": 0.0,
"step": 13525
},
{
"epoch": 57.66,
"learning_rate": 1.0027586206896553e-06,
"loss": 0.0,
"step": 13550
},
{
"epoch": 57.77,
"learning_rate": 9.855172413793104e-07,
"loss": 0.0,
"step": 13575
},
{
"epoch": 57.87,
"learning_rate": 9.682758620689656e-07,
"loss": 0.0,
"step": 13600
},
{
"epoch": 57.98,
"learning_rate": 9.510344827586207e-07,
"loss": 0.0,
"step": 13625
},
{
"epoch": 58.09,
"learning_rate": 9.33793103448276e-07,
"loss": 0.0,
"step": 13650
},
{
"epoch": 58.19,
"learning_rate": 9.165517241379311e-07,
"loss": 0.0,
"step": 13675
},
{
"epoch": 58.3,
"learning_rate": 8.993103448275863e-07,
"loss": 0.0,
"step": 13700
},
{
"epoch": 58.4,
"learning_rate": 8.820689655172414e-07,
"loss": 0.0,
"step": 13725
},
{
"epoch": 58.51,
"learning_rate": 8.648275862068967e-07,
"loss": 0.0,
"step": 13750
},
{
"epoch": 58.62,
"learning_rate": 8.475862068965517e-07,
"loss": 0.0,
"step": 13775
},
{
"epoch": 58.72,
"learning_rate": 8.303448275862069e-07,
"loss": 0.0,
"step": 13800
},
{
"epoch": 58.83,
"learning_rate": 8.131034482758621e-07,
"loss": 0.0,
"step": 13825
},
{
"epoch": 58.94,
"learning_rate": 7.958620689655173e-07,
"loss": 0.0,
"step": 13850
},
{
"epoch": 59.04,
"learning_rate": 7.786206896551725e-07,
"loss": 0.0,
"step": 13875
},
{
"epoch": 59.15,
"learning_rate": 7.613793103448276e-07,
"loss": 0.0,
"step": 13900
},
{
"epoch": 59.26,
"learning_rate": 7.441379310344828e-07,
"loss": 0.0,
"step": 13925
},
{
"epoch": 59.36,
"learning_rate": 7.268965517241381e-07,
"loss": 0.0,
"step": 13950
},
{
"epoch": 59.47,
"learning_rate": 7.096551724137931e-07,
"loss": 0.0,
"step": 13975
},
{
"epoch": 59.57,
"learning_rate": 6.924137931034483e-07,
"loss": 0.0,
"step": 14000
},
{
"epoch": 59.57,
"eval_cer": 10.923626496298967,
"eval_loss": 0.6015200018882751,
"eval_runtime": 2141.9575,
"eval_samples_per_second": 0.881,
"eval_steps_per_second": 0.055,
"eval_wer": 33.029276818876994,
"step": 14000
},
{
"epoch": 59.68,
"learning_rate": 6.751724137931034e-07,
"loss": 0.0,
"step": 14025
},
{
"epoch": 59.79,
"learning_rate": 6.579310344827587e-07,
"loss": 0.0,
"step": 14050
},
{
"epoch": 59.89,
"learning_rate": 6.406896551724139e-07,
"loss": 0.0,
"step": 14075
},
{
"epoch": 60.0,
"learning_rate": 6.23448275862069e-07,
"loss": 0.0,
"step": 14100
},
{
"epoch": 60.11,
"learning_rate": 6.062068965517242e-07,
"loss": 0.0,
"step": 14125
},
{
"epoch": 60.21,
"learning_rate": 5.889655172413794e-07,
"loss": 0.0,
"step": 14150
},
{
"epoch": 60.32,
"learning_rate": 5.717241379310345e-07,
"loss": 0.0,
"step": 14175
},
{
"epoch": 60.43,
"learning_rate": 5.544827586206897e-07,
"loss": 0.0,
"step": 14200
},
{
"epoch": 60.53,
"learning_rate": 5.372413793103449e-07,
"loss": 0.0,
"step": 14225
},
{
"epoch": 60.64,
"learning_rate": 5.2e-07,
"loss": 0.0,
"step": 14250
},
{
"epoch": 60.74,
"learning_rate": 5.027586206896552e-07,
"loss": 0.0,
"step": 14275
},
{
"epoch": 60.85,
"learning_rate": 4.855172413793104e-07,
"loss": 0.0,
"step": 14300
},
{
"epoch": 60.96,
"learning_rate": 4.6827586206896556e-07,
"loss": 0.0,
"step": 14325
},
{
"epoch": 61.06,
"learning_rate": 4.510344827586207e-07,
"loss": 0.0,
"step": 14350
},
{
"epoch": 61.17,
"learning_rate": 4.337931034482759e-07,
"loss": 0.0,
"step": 14375
},
{
"epoch": 61.28,
"learning_rate": 4.1655172413793107e-07,
"loss": 0.0,
"step": 14400
},
{
"epoch": 61.38,
"learning_rate": 3.9931034482758626e-07,
"loss": 0.0,
"step": 14425
},
{
"epoch": 61.49,
"learning_rate": 3.820689655172414e-07,
"loss": 0.0,
"step": 14450
},
{
"epoch": 61.6,
"learning_rate": 3.648275862068966e-07,
"loss": 0.0,
"step": 14475
},
{
"epoch": 61.7,
"learning_rate": 3.475862068965517e-07,
"loss": 0.0,
"step": 14500
},
{
"epoch": 61.81,
"learning_rate": 3.3034482758620696e-07,
"loss": 0.0,
"step": 14525
},
{
"epoch": 61.91,
"learning_rate": 3.131034482758621e-07,
"loss": 0.0,
"step": 14550
},
{
"epoch": 62.02,
"learning_rate": 2.958620689655173e-07,
"loss": 0.0,
"step": 14575
},
{
"epoch": 62.13,
"learning_rate": 2.7862068965517247e-07,
"loss": 0.0,
"step": 14600
},
{
"epoch": 62.23,
"learning_rate": 2.613793103448276e-07,
"loss": 0.0,
"step": 14625
},
{
"epoch": 62.34,
"learning_rate": 2.441379310344828e-07,
"loss": 0.0,
"step": 14650
},
{
"epoch": 62.45,
"learning_rate": 2.2689655172413795e-07,
"loss": 0.0,
"step": 14675
},
{
"epoch": 62.55,
"learning_rate": 2.0965517241379314e-07,
"loss": 0.0,
"step": 14700
},
{
"epoch": 62.66,
"learning_rate": 1.924137931034483e-07,
"loss": 0.0,
"step": 14725
},
{
"epoch": 62.77,
"learning_rate": 1.751724137931035e-07,
"loss": 0.0,
"step": 14750
},
{
"epoch": 62.87,
"learning_rate": 1.5793103448275865e-07,
"loss": 0.0,
"step": 14775
},
{
"epoch": 62.98,
"learning_rate": 1.406896551724138e-07,
"loss": 0.0,
"step": 14800
},
{
"epoch": 63.09,
"learning_rate": 1.2344827586206897e-07,
"loss": 0.0,
"step": 14825
},
{
"epoch": 63.19,
"learning_rate": 1.0620689655172415e-07,
"loss": 0.0,
"step": 14850
},
{
"epoch": 63.3,
"learning_rate": 8.896551724137932e-08,
"loss": 0.0,
"step": 14875
},
{
"epoch": 63.4,
"learning_rate": 7.172413793103448e-08,
"loss": 0.0,
"step": 14900
},
{
"epoch": 63.51,
"learning_rate": 5.448275862068966e-08,
"loss": 0.0,
"step": 14925
},
{
"epoch": 63.62,
"learning_rate": 3.724137931034483e-08,
"loss": 0.0,
"step": 14950
},
{
"epoch": 63.72,
"learning_rate": 2e-08,
"loss": 0.0,
"step": 14975
},
{
"epoch": 63.83,
"learning_rate": 2.758620689655173e-09,
"loss": 0.0,
"step": 15000
},
{
"epoch": 63.83,
"eval_cer": 10.912463290226183,
"eval_loss": 0.6045265197753906,
"eval_runtime": 2136.6496,
"eval_samples_per_second": 0.883,
"eval_steps_per_second": 0.055,
"eval_wer": 33.03473891195106,
"step": 15000
},
{
"epoch": 63.83,
"step": 15000,
"total_flos": 4.884757972077773e+20,
"train_loss": 0.04087234922135443,
"train_runtime": 89952.8574,
"train_samples_per_second": 5.336,
"train_steps_per_second": 0.167
},
{
"epoch": 63.83,
"eval_cer": 10.923626496298967,
"eval_loss": 0.6015200018882751,
"eval_runtime": 2136.8751,
"eval_samples_per_second": 0.883,
"eval_steps_per_second": 0.055,
"eval_wer": 33.029276818876994,
"step": 15000
}
],
"max_steps": 15000,
"num_train_epochs": 64,
"total_flos": 4.884757972077773e+20,
"trial_name": null,
"trial_params": null
}