whisper-sm-el-intlv-xl / trainer_state.json
farsipal's picture
Upload 21 files
b489a3b
{
"best_metric": 19.48365527488856,
"best_model_checkpoint": "./data/finetuningRuns/whisper-sm-el-intlv-xl/checkpoint-6000",
"epoch": 24.875621890547265,
"global_step": 10000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.06,
"learning_rate": 4.08e-07,
"loss": 0.9083,
"step": 25
},
{
"epoch": 0.12,
"learning_rate": 8.33e-07,
"loss": 0.8372,
"step": 50
},
{
"epoch": 0.19,
"learning_rate": 1.2579999999999999e-06,
"loss": 0.661,
"step": 75
},
{
"epoch": 0.25,
"learning_rate": 1.683e-06,
"loss": 0.5264,
"step": 100
},
{
"epoch": 0.31,
"learning_rate": 2.108e-06,
"loss": 0.4432,
"step": 125
},
{
"epoch": 0.37,
"learning_rate": 2.5329999999999997e-06,
"loss": 0.3872,
"step": 150
},
{
"epoch": 0.44,
"learning_rate": 2.958e-06,
"loss": 0.3603,
"step": 175
},
{
"epoch": 0.5,
"learning_rate": 3.383e-06,
"loss": 0.3287,
"step": 200
},
{
"epoch": 0.56,
"learning_rate": 3.8080000000000002e-06,
"loss": 0.3287,
"step": 225
},
{
"epoch": 0.62,
"learning_rate": 4.233e-06,
"loss": 0.2921,
"step": 250
},
{
"epoch": 0.68,
"learning_rate": 4.658e-06,
"loss": 0.2808,
"step": 275
},
{
"epoch": 0.75,
"learning_rate": 5.0829999999999994e-06,
"loss": 0.2582,
"step": 300
},
{
"epoch": 0.81,
"learning_rate": 5.5080000000000005e-06,
"loss": 0.2577,
"step": 325
},
{
"epoch": 0.87,
"learning_rate": 5.933e-06,
"loss": 0.2357,
"step": 350
},
{
"epoch": 0.93,
"learning_rate": 6.358e-06,
"loss": 0.2202,
"step": 375
},
{
"epoch": 1.0,
"learning_rate": 6.783e-06,
"loss": 0.2263,
"step": 400
},
{
"epoch": 1.06,
"learning_rate": 7.2079999999999994e-06,
"loss": 0.1766,
"step": 425
},
{
"epoch": 1.12,
"learning_rate": 7.633e-06,
"loss": 0.173,
"step": 450
},
{
"epoch": 1.18,
"learning_rate": 8.058e-06,
"loss": 0.1568,
"step": 475
},
{
"epoch": 1.24,
"learning_rate": 8.483e-06,
"loss": 0.159,
"step": 500
},
{
"epoch": 1.31,
"learning_rate": 8.478526315789473e-06,
"loss": 0.1435,
"step": 525
},
{
"epoch": 1.37,
"learning_rate": 8.456157894736841e-06,
"loss": 0.1501,
"step": 550
},
{
"epoch": 1.43,
"learning_rate": 8.43378947368421e-06,
"loss": 0.1509,
"step": 575
},
{
"epoch": 1.49,
"learning_rate": 8.41142105263158e-06,
"loss": 0.1538,
"step": 600
},
{
"epoch": 1.55,
"learning_rate": 8.389052631578948e-06,
"loss": 0.1472,
"step": 625
},
{
"epoch": 1.62,
"learning_rate": 8.366684210526316e-06,
"loss": 0.1302,
"step": 650
},
{
"epoch": 1.68,
"learning_rate": 8.344315789473684e-06,
"loss": 0.1291,
"step": 675
},
{
"epoch": 1.74,
"learning_rate": 8.321947368421052e-06,
"loss": 0.1396,
"step": 700
},
{
"epoch": 1.8,
"learning_rate": 8.29957894736842e-06,
"loss": 0.1175,
"step": 725
},
{
"epoch": 1.87,
"learning_rate": 8.277210526315789e-06,
"loss": 0.1187,
"step": 750
},
{
"epoch": 1.93,
"learning_rate": 8.254842105263159e-06,
"loss": 0.1285,
"step": 775
},
{
"epoch": 1.99,
"learning_rate": 8.232473684210527e-06,
"loss": 0.1168,
"step": 800
},
{
"epoch": 2.05,
"learning_rate": 8.210105263157895e-06,
"loss": 0.0795,
"step": 825
},
{
"epoch": 2.11,
"learning_rate": 8.187736842105263e-06,
"loss": 0.0612,
"step": 850
},
{
"epoch": 2.18,
"learning_rate": 8.165368421052632e-06,
"loss": 0.058,
"step": 875
},
{
"epoch": 2.24,
"learning_rate": 8.143e-06,
"loss": 0.0644,
"step": 900
},
{
"epoch": 2.3,
"learning_rate": 8.120631578947368e-06,
"loss": 0.0653,
"step": 925
},
{
"epoch": 2.36,
"learning_rate": 8.098263157894736e-06,
"loss": 0.0678,
"step": 950
},
{
"epoch": 2.43,
"learning_rate": 8.075894736842106e-06,
"loss": 0.0546,
"step": 975
},
{
"epoch": 2.49,
"learning_rate": 8.053526315789474e-06,
"loss": 0.0545,
"step": 1000
},
{
"epoch": 2.49,
"eval_loss": 0.28911715745925903,
"eval_runtime": 1701.0873,
"eval_samples_per_second": 0.997,
"eval_steps_per_second": 0.125,
"eval_wer": 22.4925705794948,
"step": 1000
},
{
"epoch": 2.55,
"learning_rate": 8.031157894736843e-06,
"loss": 0.0532,
"step": 1025
},
{
"epoch": 2.61,
"learning_rate": 8.008789473684211e-06,
"loss": 0.0656,
"step": 1050
},
{
"epoch": 2.67,
"learning_rate": 7.986421052631579e-06,
"loss": 0.056,
"step": 1075
},
{
"epoch": 2.74,
"learning_rate": 7.964052631578947e-06,
"loss": 0.0573,
"step": 1100
},
{
"epoch": 2.8,
"learning_rate": 7.941684210526316e-06,
"loss": 0.0509,
"step": 1125
},
{
"epoch": 2.86,
"learning_rate": 7.919315789473684e-06,
"loss": 0.0572,
"step": 1150
},
{
"epoch": 2.92,
"learning_rate": 7.896947368421052e-06,
"loss": 0.0516,
"step": 1175
},
{
"epoch": 2.99,
"learning_rate": 7.87457894736842e-06,
"loss": 0.051,
"step": 1200
},
{
"epoch": 3.05,
"learning_rate": 7.852210526315789e-06,
"loss": 0.0306,
"step": 1225
},
{
"epoch": 3.11,
"learning_rate": 7.829842105263157e-06,
"loss": 0.0236,
"step": 1250
},
{
"epoch": 3.17,
"learning_rate": 7.807473684210525e-06,
"loss": 0.0239,
"step": 1275
},
{
"epoch": 3.23,
"learning_rate": 7.785105263157895e-06,
"loss": 0.0219,
"step": 1300
},
{
"epoch": 3.3,
"learning_rate": 7.762736842105263e-06,
"loss": 0.0245,
"step": 1325
},
{
"epoch": 3.36,
"learning_rate": 7.740368421052631e-06,
"loss": 0.0241,
"step": 1350
},
{
"epoch": 3.42,
"learning_rate": 7.718e-06,
"loss": 0.0222,
"step": 1375
},
{
"epoch": 3.48,
"learning_rate": 7.695631578947368e-06,
"loss": 0.0223,
"step": 1400
},
{
"epoch": 3.54,
"learning_rate": 7.673263157894736e-06,
"loss": 0.028,
"step": 1425
},
{
"epoch": 3.61,
"learning_rate": 7.650894736842106e-06,
"loss": 0.0251,
"step": 1450
},
{
"epoch": 3.67,
"learning_rate": 7.628526315789473e-06,
"loss": 0.0242,
"step": 1475
},
{
"epoch": 3.73,
"learning_rate": 7.606157894736842e-06,
"loss": 0.0232,
"step": 1500
},
{
"epoch": 3.79,
"learning_rate": 7.58378947368421e-06,
"loss": 0.0213,
"step": 1525
},
{
"epoch": 3.86,
"learning_rate": 7.561421052631579e-06,
"loss": 0.0199,
"step": 1550
},
{
"epoch": 3.92,
"learning_rate": 7.539052631578947e-06,
"loss": 0.022,
"step": 1575
},
{
"epoch": 3.98,
"learning_rate": 7.5166842105263154e-06,
"loss": 0.0216,
"step": 1600
},
{
"epoch": 4.04,
"learning_rate": 7.494315789473684e-06,
"loss": 0.014,
"step": 1625
},
{
"epoch": 4.1,
"learning_rate": 7.471947368421053e-06,
"loss": 0.0088,
"step": 1650
},
{
"epoch": 4.17,
"learning_rate": 7.449578947368421e-06,
"loss": 0.0093,
"step": 1675
},
{
"epoch": 4.23,
"learning_rate": 7.42721052631579e-06,
"loss": 0.0115,
"step": 1700
},
{
"epoch": 4.29,
"learning_rate": 7.404842105263158e-06,
"loss": 0.0085,
"step": 1725
},
{
"epoch": 4.35,
"learning_rate": 7.3824736842105265e-06,
"loss": 0.0103,
"step": 1750
},
{
"epoch": 4.42,
"learning_rate": 7.360105263157895e-06,
"loss": 0.009,
"step": 1775
},
{
"epoch": 4.48,
"learning_rate": 7.337736842105263e-06,
"loss": 0.0104,
"step": 1800
},
{
"epoch": 4.54,
"learning_rate": 7.315368421052631e-06,
"loss": 0.0089,
"step": 1825
},
{
"epoch": 4.6,
"learning_rate": 7.2929999999999995e-06,
"loss": 0.0106,
"step": 1850
},
{
"epoch": 4.66,
"learning_rate": 7.2706315789473685e-06,
"loss": 0.0095,
"step": 1875
},
{
"epoch": 4.73,
"learning_rate": 7.248263157894737e-06,
"loss": 0.009,
"step": 1900
},
{
"epoch": 4.79,
"learning_rate": 7.225894736842105e-06,
"loss": 0.0081,
"step": 1925
},
{
"epoch": 4.85,
"learning_rate": 7.203526315789473e-06,
"loss": 0.0099,
"step": 1950
},
{
"epoch": 4.91,
"learning_rate": 7.181157894736842e-06,
"loss": 0.0103,
"step": 1975
},
{
"epoch": 4.98,
"learning_rate": 7.1587894736842106e-06,
"loss": 0.0093,
"step": 2000
},
{
"epoch": 4.98,
"eval_loss": 0.39272043108940125,
"eval_runtime": 1655.9342,
"eval_samples_per_second": 1.024,
"eval_steps_per_second": 0.128,
"eval_wer": 20.13372956909361,
"step": 2000
},
{
"epoch": 5.04,
"learning_rate": 7.13642105263158e-06,
"loss": 0.0074,
"step": 2025
},
{
"epoch": 5.1,
"learning_rate": 7.114052631578948e-06,
"loss": 0.0046,
"step": 2050
},
{
"epoch": 5.16,
"learning_rate": 7.091684210526316e-06,
"loss": 0.0056,
"step": 2075
},
{
"epoch": 5.22,
"learning_rate": 7.069315789473684e-06,
"loss": 0.0051,
"step": 2100
},
{
"epoch": 5.29,
"learning_rate": 7.0469473684210526e-06,
"loss": 0.0048,
"step": 2125
},
{
"epoch": 5.35,
"learning_rate": 7.024578947368421e-06,
"loss": 0.0045,
"step": 2150
},
{
"epoch": 5.41,
"learning_rate": 7.002210526315789e-06,
"loss": 0.0041,
"step": 2175
},
{
"epoch": 5.47,
"learning_rate": 6.979842105263157e-06,
"loss": 0.004,
"step": 2200
},
{
"epoch": 5.53,
"learning_rate": 6.957473684210526e-06,
"loss": 0.0041,
"step": 2225
},
{
"epoch": 5.6,
"learning_rate": 6.935105263157895e-06,
"loss": 0.0039,
"step": 2250
},
{
"epoch": 5.66,
"learning_rate": 6.912736842105263e-06,
"loss": 0.0043,
"step": 2275
},
{
"epoch": 5.72,
"learning_rate": 6.890368421052631e-06,
"loss": 0.0038,
"step": 2300
},
{
"epoch": 5.78,
"learning_rate": 6.868e-06,
"loss": 0.0045,
"step": 2325
},
{
"epoch": 5.85,
"learning_rate": 6.845631578947368e-06,
"loss": 0.0042,
"step": 2350
},
{
"epoch": 5.91,
"learning_rate": 6.8232631578947375e-06,
"loss": 0.0056,
"step": 2375
},
{
"epoch": 5.97,
"learning_rate": 6.800894736842106e-06,
"loss": 0.0049,
"step": 2400
},
{
"epoch": 6.03,
"learning_rate": 6.778526315789474e-06,
"loss": 0.004,
"step": 2425
},
{
"epoch": 6.09,
"learning_rate": 6.756157894736842e-06,
"loss": 0.0029,
"step": 2450
},
{
"epoch": 6.16,
"learning_rate": 6.73378947368421e-06,
"loss": 0.0028,
"step": 2475
},
{
"epoch": 6.22,
"learning_rate": 6.711421052631579e-06,
"loss": 0.0029,
"step": 2500
},
{
"epoch": 6.28,
"learning_rate": 6.689052631578947e-06,
"loss": 0.003,
"step": 2525
},
{
"epoch": 6.34,
"learning_rate": 6.666684210526316e-06,
"loss": 0.003,
"step": 2550
},
{
"epoch": 6.41,
"learning_rate": 6.644315789473684e-06,
"loss": 0.0033,
"step": 2575
},
{
"epoch": 6.47,
"learning_rate": 6.621947368421052e-06,
"loss": 0.0028,
"step": 2600
},
{
"epoch": 6.53,
"learning_rate": 6.599578947368421e-06,
"loss": 0.0036,
"step": 2625
},
{
"epoch": 6.59,
"learning_rate": 6.577210526315789e-06,
"loss": 0.0028,
"step": 2650
},
{
"epoch": 6.65,
"learning_rate": 6.554842105263158e-06,
"loss": 0.0024,
"step": 2675
},
{
"epoch": 6.72,
"learning_rate": 6.532473684210526e-06,
"loss": 0.0028,
"step": 2700
},
{
"epoch": 6.78,
"learning_rate": 6.510105263157895e-06,
"loss": 0.0024,
"step": 2725
},
{
"epoch": 6.84,
"learning_rate": 6.4877368421052635e-06,
"loss": 0.0027,
"step": 2750
},
{
"epoch": 6.9,
"learning_rate": 6.465368421052632e-06,
"loss": 0.0025,
"step": 2775
},
{
"epoch": 6.97,
"learning_rate": 6.443e-06,
"loss": 0.0029,
"step": 2800
},
{
"epoch": 7.03,
"learning_rate": 6.420631578947368e-06,
"loss": 0.0022,
"step": 2825
},
{
"epoch": 7.09,
"learning_rate": 6.3982631578947364e-06,
"loss": 0.0022,
"step": 2850
},
{
"epoch": 7.15,
"learning_rate": 6.375894736842105e-06,
"loss": 0.002,
"step": 2875
},
{
"epoch": 7.21,
"learning_rate": 6.353526315789474e-06,
"loss": 0.0019,
"step": 2900
},
{
"epoch": 7.28,
"learning_rate": 6.331157894736842e-06,
"loss": 0.0021,
"step": 2925
},
{
"epoch": 7.34,
"learning_rate": 6.30878947368421e-06,
"loss": 0.0017,
"step": 2950
},
{
"epoch": 7.4,
"learning_rate": 6.2864210526315785e-06,
"loss": 0.0016,
"step": 2975
},
{
"epoch": 7.46,
"learning_rate": 6.2640526315789475e-06,
"loss": 0.0018,
"step": 3000
},
{
"epoch": 7.46,
"eval_loss": 0.40309956669807434,
"eval_runtime": 1128.4825,
"eval_samples_per_second": 1.503,
"eval_steps_per_second": 0.188,
"eval_wer": 20.161589895988115,
"step": 3000
},
{
"epoch": 7.52,
"learning_rate": 6.241684210526316e-06,
"loss": 0.0019,
"step": 3025
},
{
"epoch": 7.59,
"learning_rate": 6.219315789473685e-06,
"loss": 0.0014,
"step": 3050
},
{
"epoch": 7.65,
"learning_rate": 6.196947368421053e-06,
"loss": 0.0014,
"step": 3075
},
{
"epoch": 7.71,
"learning_rate": 6.174578947368421e-06,
"loss": 0.0014,
"step": 3100
},
{
"epoch": 7.77,
"learning_rate": 6.1522105263157896e-06,
"loss": 0.0017,
"step": 3125
},
{
"epoch": 7.84,
"learning_rate": 6.129842105263158e-06,
"loss": 0.0019,
"step": 3150
},
{
"epoch": 7.9,
"learning_rate": 6.107473684210526e-06,
"loss": 0.0016,
"step": 3175
},
{
"epoch": 7.96,
"learning_rate": 6.085105263157894e-06,
"loss": 0.0017,
"step": 3200
},
{
"epoch": 8.02,
"learning_rate": 6.062736842105263e-06,
"loss": 0.0018,
"step": 3225
},
{
"epoch": 8.08,
"learning_rate": 6.0403684210526316e-06,
"loss": 0.0014,
"step": 3250
},
{
"epoch": 8.15,
"learning_rate": 6.018e-06,
"loss": 0.0013,
"step": 3275
},
{
"epoch": 8.21,
"learning_rate": 5.995631578947368e-06,
"loss": 0.0012,
"step": 3300
},
{
"epoch": 8.27,
"learning_rate": 5.973263157894736e-06,
"loss": 0.0017,
"step": 3325
},
{
"epoch": 8.33,
"learning_rate": 5.950894736842105e-06,
"loss": 0.0016,
"step": 3350
},
{
"epoch": 8.4,
"learning_rate": 5.928526315789474e-06,
"loss": 0.0019,
"step": 3375
},
{
"epoch": 8.46,
"learning_rate": 5.906157894736843e-06,
"loss": 0.0013,
"step": 3400
},
{
"epoch": 8.52,
"learning_rate": 5.883789473684211e-06,
"loss": 0.0014,
"step": 3425
},
{
"epoch": 8.58,
"learning_rate": 5.861421052631579e-06,
"loss": 0.0021,
"step": 3450
},
{
"epoch": 8.64,
"learning_rate": 5.839052631578947e-06,
"loss": 0.0014,
"step": 3475
},
{
"epoch": 8.71,
"learning_rate": 5.816684210526316e-06,
"loss": 0.0012,
"step": 3500
},
{
"epoch": 8.77,
"learning_rate": 5.794315789473684e-06,
"loss": 0.0013,
"step": 3525
},
{
"epoch": 8.83,
"learning_rate": 5.771947368421052e-06,
"loss": 0.0013,
"step": 3550
},
{
"epoch": 8.89,
"learning_rate": 5.749578947368421e-06,
"loss": 0.0014,
"step": 3575
},
{
"epoch": 8.96,
"learning_rate": 5.727210526315789e-06,
"loss": 0.0017,
"step": 3600
},
{
"epoch": 9.02,
"learning_rate": 5.704842105263158e-06,
"loss": 0.0018,
"step": 3625
},
{
"epoch": 9.08,
"learning_rate": 5.682473684210526e-06,
"loss": 0.001,
"step": 3650
},
{
"epoch": 9.14,
"learning_rate": 5.660105263157895e-06,
"loss": 0.0015,
"step": 3675
},
{
"epoch": 9.2,
"learning_rate": 5.637736842105263e-06,
"loss": 0.0017,
"step": 3700
},
{
"epoch": 9.27,
"learning_rate": 5.615368421052632e-06,
"loss": 0.0018,
"step": 3725
},
{
"epoch": 9.33,
"learning_rate": 5.5930000000000005e-06,
"loss": 0.0028,
"step": 3750
},
{
"epoch": 9.39,
"learning_rate": 5.570631578947369e-06,
"loss": 0.0015,
"step": 3775
},
{
"epoch": 9.45,
"learning_rate": 5.548263157894737e-06,
"loss": 0.0014,
"step": 3800
},
{
"epoch": 9.51,
"learning_rate": 5.525894736842105e-06,
"loss": 0.0012,
"step": 3825
},
{
"epoch": 9.58,
"learning_rate": 5.5035263157894734e-06,
"loss": 0.0015,
"step": 3850
},
{
"epoch": 9.64,
"learning_rate": 5.481157894736842e-06,
"loss": 0.001,
"step": 3875
},
{
"epoch": 9.7,
"learning_rate": 5.45878947368421e-06,
"loss": 0.0014,
"step": 3900
},
{
"epoch": 9.76,
"learning_rate": 5.436421052631579e-06,
"loss": 0.0014,
"step": 3925
},
{
"epoch": 9.83,
"learning_rate": 5.414052631578947e-06,
"loss": 0.0025,
"step": 3950
},
{
"epoch": 9.89,
"learning_rate": 5.3916842105263154e-06,
"loss": 0.0016,
"step": 3975
},
{
"epoch": 9.95,
"learning_rate": 5.369315789473684e-06,
"loss": 0.001,
"step": 4000
},
{
"epoch": 9.95,
"eval_loss": 0.4208928644657135,
"eval_runtime": 1115.4906,
"eval_samples_per_second": 1.52,
"eval_steps_per_second": 0.19,
"eval_wer": 19.687964338781576,
"step": 4000
},
{
"epoch": 10.01,
"learning_rate": 5.346947368421053e-06,
"loss": 0.0011,
"step": 4025
},
{
"epoch": 10.07,
"learning_rate": 5.324578947368421e-06,
"loss": 0.0011,
"step": 4050
},
{
"epoch": 10.14,
"learning_rate": 5.30221052631579e-06,
"loss": 0.0008,
"step": 4075
},
{
"epoch": 10.2,
"learning_rate": 5.279842105263158e-06,
"loss": 0.002,
"step": 4100
},
{
"epoch": 10.26,
"learning_rate": 5.2574736842105265e-06,
"loss": 0.0011,
"step": 4125
},
{
"epoch": 10.32,
"learning_rate": 5.235105263157895e-06,
"loss": 0.0007,
"step": 4150
},
{
"epoch": 10.39,
"learning_rate": 5.212736842105263e-06,
"loss": 0.0009,
"step": 4175
},
{
"epoch": 10.45,
"learning_rate": 5.190368421052631e-06,
"loss": 0.0009,
"step": 4200
},
{
"epoch": 10.51,
"learning_rate": 5.1679999999999995e-06,
"loss": 0.001,
"step": 4225
},
{
"epoch": 10.57,
"learning_rate": 5.1456315789473686e-06,
"loss": 0.0011,
"step": 4250
},
{
"epoch": 10.63,
"learning_rate": 5.123263157894737e-06,
"loss": 0.0009,
"step": 4275
},
{
"epoch": 10.7,
"learning_rate": 5.100894736842105e-06,
"loss": 0.0008,
"step": 4300
},
{
"epoch": 10.76,
"learning_rate": 5.078526315789473e-06,
"loss": 0.0007,
"step": 4325
},
{
"epoch": 10.82,
"learning_rate": 5.056157894736842e-06,
"loss": 0.0007,
"step": 4350
},
{
"epoch": 10.88,
"learning_rate": 5.0337894736842106e-06,
"loss": 0.0007,
"step": 4375
},
{
"epoch": 10.95,
"learning_rate": 5.01142105263158e-06,
"loss": 0.0007,
"step": 4400
},
{
"epoch": 11.01,
"learning_rate": 4.989052631578948e-06,
"loss": 0.0008,
"step": 4425
},
{
"epoch": 11.07,
"learning_rate": 4.966684210526316e-06,
"loss": 0.0007,
"step": 4450
},
{
"epoch": 11.13,
"learning_rate": 4.944315789473684e-06,
"loss": 0.0008,
"step": 4475
},
{
"epoch": 11.19,
"learning_rate": 4.921947368421053e-06,
"loss": 0.0005,
"step": 4500
},
{
"epoch": 11.26,
"learning_rate": 4.899578947368421e-06,
"loss": 0.0012,
"step": 4525
},
{
"epoch": 11.32,
"learning_rate": 4.877210526315789e-06,
"loss": 0.0005,
"step": 4550
},
{
"epoch": 11.38,
"learning_rate": 4.854842105263157e-06,
"loss": 0.0005,
"step": 4575
},
{
"epoch": 11.44,
"learning_rate": 4.832473684210526e-06,
"loss": 0.0015,
"step": 4600
},
{
"epoch": 11.5,
"learning_rate": 4.810105263157895e-06,
"loss": 0.001,
"step": 4625
},
{
"epoch": 11.57,
"learning_rate": 4.788631578947368e-06,
"loss": 0.0009,
"step": 4650
},
{
"epoch": 11.63,
"learning_rate": 4.7662631578947375e-06,
"loss": 0.0012,
"step": 4675
},
{
"epoch": 11.69,
"learning_rate": 4.743894736842106e-06,
"loss": 0.0006,
"step": 4700
},
{
"epoch": 11.75,
"learning_rate": 4.721526315789474e-06,
"loss": 0.0009,
"step": 4725
},
{
"epoch": 11.82,
"learning_rate": 4.699157894736842e-06,
"loss": 0.0007,
"step": 4750
},
{
"epoch": 11.88,
"learning_rate": 4.6767894736842104e-06,
"loss": 0.0006,
"step": 4775
},
{
"epoch": 11.94,
"learning_rate": 4.654421052631579e-06,
"loss": 0.001,
"step": 4800
},
{
"epoch": 12.0,
"learning_rate": 4.632052631578947e-06,
"loss": 0.0008,
"step": 4825
},
{
"epoch": 12.06,
"learning_rate": 4.609684210526316e-06,
"loss": 0.0013,
"step": 4850
},
{
"epoch": 12.13,
"learning_rate": 4.587315789473684e-06,
"loss": 0.0006,
"step": 4875
},
{
"epoch": 12.19,
"learning_rate": 4.5649473684210525e-06,
"loss": 0.0008,
"step": 4900
},
{
"epoch": 12.25,
"learning_rate": 4.542578947368421e-06,
"loss": 0.0007,
"step": 4925
},
{
"epoch": 12.31,
"learning_rate": 4.520210526315789e-06,
"loss": 0.0006,
"step": 4950
},
{
"epoch": 12.38,
"learning_rate": 4.497842105263158e-06,
"loss": 0.0011,
"step": 4975
},
{
"epoch": 12.44,
"learning_rate": 4.475473684210527e-06,
"loss": 0.0008,
"step": 5000
},
{
"epoch": 12.44,
"eval_loss": 0.4497528672218323,
"eval_runtime": 1118.1052,
"eval_samples_per_second": 1.517,
"eval_steps_per_second": 0.19,
"eval_wer": 20.096582466567607,
"step": 5000
},
{
"epoch": 12.5,
"learning_rate": 4.453105263157895e-06,
"loss": 0.0007,
"step": 5025
},
{
"epoch": 12.56,
"learning_rate": 4.4307368421052636e-06,
"loss": 0.0008,
"step": 5050
},
{
"epoch": 12.62,
"learning_rate": 4.408368421052632e-06,
"loss": 0.0011,
"step": 5075
},
{
"epoch": 12.69,
"learning_rate": 4.386e-06,
"loss": 0.0011,
"step": 5100
},
{
"epoch": 12.75,
"learning_rate": 4.363631578947368e-06,
"loss": 0.0008,
"step": 5125
},
{
"epoch": 12.81,
"learning_rate": 4.3412631578947365e-06,
"loss": 0.0005,
"step": 5150
},
{
"epoch": 12.87,
"learning_rate": 4.318894736842105e-06,
"loss": 0.0006,
"step": 5175
},
{
"epoch": 12.94,
"learning_rate": 4.296526315789474e-06,
"loss": 0.0005,
"step": 5200
},
{
"epoch": 13.0,
"learning_rate": 4.274157894736842e-06,
"loss": 0.0008,
"step": 5225
},
{
"epoch": 13.06,
"learning_rate": 4.25178947368421e-06,
"loss": 0.0005,
"step": 5250
},
{
"epoch": 13.12,
"learning_rate": 4.229421052631579e-06,
"loss": 0.0005,
"step": 5275
},
{
"epoch": 13.18,
"learning_rate": 4.207052631578948e-06,
"loss": 0.0004,
"step": 5300
},
{
"epoch": 13.25,
"learning_rate": 4.184684210526316e-06,
"loss": 0.0005,
"step": 5325
},
{
"epoch": 13.31,
"learning_rate": 4.162315789473684e-06,
"loss": 0.0008,
"step": 5350
},
{
"epoch": 13.37,
"learning_rate": 4.139947368421052e-06,
"loss": 0.0005,
"step": 5375
},
{
"epoch": 13.43,
"learning_rate": 4.117578947368421e-06,
"loss": 0.0006,
"step": 5400
},
{
"epoch": 13.5,
"learning_rate": 4.09521052631579e-06,
"loss": 0.0005,
"step": 5425
},
{
"epoch": 13.56,
"learning_rate": 4.072842105263158e-06,
"loss": 0.0005,
"step": 5450
},
{
"epoch": 13.62,
"learning_rate": 4.050473684210526e-06,
"loss": 0.0012,
"step": 5475
},
{
"epoch": 13.68,
"learning_rate": 4.028105263157894e-06,
"loss": 0.0009,
"step": 5500
},
{
"epoch": 13.74,
"learning_rate": 4.005736842105263e-06,
"loss": 0.0008,
"step": 5525
},
{
"epoch": 13.81,
"learning_rate": 3.983368421052632e-06,
"loss": 0.001,
"step": 5550
},
{
"epoch": 13.87,
"learning_rate": 3.961e-06,
"loss": 0.0009,
"step": 5575
},
{
"epoch": 13.93,
"learning_rate": 3.938631578947369e-06,
"loss": 0.0012,
"step": 5600
},
{
"epoch": 13.99,
"learning_rate": 3.916263157894737e-06,
"loss": 0.0007,
"step": 5625
},
{
"epoch": 14.05,
"learning_rate": 3.893894736842105e-06,
"loss": 0.0006,
"step": 5650
},
{
"epoch": 14.12,
"learning_rate": 3.871526315789474e-06,
"loss": 0.0006,
"step": 5675
},
{
"epoch": 14.18,
"learning_rate": 3.849157894736842e-06,
"loss": 0.0005,
"step": 5700
},
{
"epoch": 14.24,
"learning_rate": 3.82678947368421e-06,
"loss": 0.0006,
"step": 5725
},
{
"epoch": 14.3,
"learning_rate": 3.804421052631579e-06,
"loss": 0.0006,
"step": 5750
},
{
"epoch": 14.37,
"learning_rate": 3.7820526315789474e-06,
"loss": 0.0007,
"step": 5775
},
{
"epoch": 14.43,
"learning_rate": 3.7596842105263157e-06,
"loss": 0.0006,
"step": 5800
},
{
"epoch": 14.49,
"learning_rate": 3.7373157894736843e-06,
"loss": 0.001,
"step": 5825
},
{
"epoch": 14.55,
"learning_rate": 3.7149473684210525e-06,
"loss": 0.0006,
"step": 5850
},
{
"epoch": 14.61,
"learning_rate": 3.6925789473684208e-06,
"loss": 0.0007,
"step": 5875
},
{
"epoch": 14.68,
"learning_rate": 3.6702105263157894e-06,
"loss": 0.0006,
"step": 5900
},
{
"epoch": 14.74,
"learning_rate": 3.647842105263158e-06,
"loss": 0.0008,
"step": 5925
},
{
"epoch": 14.8,
"learning_rate": 3.6254736842105263e-06,
"loss": 0.0006,
"step": 5950
},
{
"epoch": 14.86,
"learning_rate": 3.6031052631578946e-06,
"loss": 0.0006,
"step": 5975
},
{
"epoch": 14.93,
"learning_rate": 3.5807368421052632e-06,
"loss": 0.0005,
"step": 6000
},
{
"epoch": 14.93,
"eval_loss": 0.47249987721443176,
"eval_runtime": 1111.7599,
"eval_samples_per_second": 1.526,
"eval_steps_per_second": 0.191,
"eval_wer": 19.48365527488856,
"step": 6000
},
{
"epoch": 14.99,
"learning_rate": 3.5583684210526315e-06,
"loss": 0.0005,
"step": 6025
},
{
"epoch": 15.05,
"learning_rate": 3.5359999999999997e-06,
"loss": 0.0006,
"step": 6050
},
{
"epoch": 15.11,
"learning_rate": 3.5136315789473683e-06,
"loss": 0.0003,
"step": 6075
},
{
"epoch": 15.17,
"learning_rate": 3.491263157894737e-06,
"loss": 0.0004,
"step": 6100
},
{
"epoch": 15.24,
"learning_rate": 3.4688947368421052e-06,
"loss": 0.0005,
"step": 6125
},
{
"epoch": 15.3,
"learning_rate": 3.446526315789474e-06,
"loss": 0.0005,
"step": 6150
},
{
"epoch": 15.36,
"learning_rate": 3.424157894736842e-06,
"loss": 0.0004,
"step": 6175
},
{
"epoch": 15.42,
"learning_rate": 3.4017894736842104e-06,
"loss": 0.0003,
"step": 6200
},
{
"epoch": 15.49,
"learning_rate": 3.3794210526315786e-06,
"loss": 0.0004,
"step": 6225
},
{
"epoch": 15.55,
"learning_rate": 3.3570526315789477e-06,
"loss": 0.0005,
"step": 6250
},
{
"epoch": 15.61,
"learning_rate": 3.334684210526316e-06,
"loss": 0.0005,
"step": 6275
},
{
"epoch": 15.67,
"learning_rate": 3.312315789473684e-06,
"loss": 0.0003,
"step": 6300
},
{
"epoch": 15.73,
"learning_rate": 3.289947368421053e-06,
"loss": 0.0003,
"step": 6325
},
{
"epoch": 15.8,
"learning_rate": 3.267578947368421e-06,
"loss": 0.0003,
"step": 6350
},
{
"epoch": 15.86,
"learning_rate": 3.2452105263157893e-06,
"loss": 0.0003,
"step": 6375
},
{
"epoch": 15.92,
"learning_rate": 3.2228421052631575e-06,
"loss": 0.0003,
"step": 6400
},
{
"epoch": 15.98,
"learning_rate": 3.2004736842105266e-06,
"loss": 0.0003,
"step": 6425
},
{
"epoch": 16.04,
"learning_rate": 3.178105263157895e-06,
"loss": 0.0002,
"step": 6450
},
{
"epoch": 16.11,
"learning_rate": 3.155736842105263e-06,
"loss": 0.0002,
"step": 6475
},
{
"epoch": 16.17,
"learning_rate": 3.1333684210526317e-06,
"loss": 0.0002,
"step": 6500
},
{
"epoch": 16.23,
"learning_rate": 3.111e-06,
"loss": 0.0002,
"step": 6525
},
{
"epoch": 16.29,
"learning_rate": 3.088631578947368e-06,
"loss": 0.0003,
"step": 6550
},
{
"epoch": 16.36,
"learning_rate": 3.0662631578947364e-06,
"loss": 0.0002,
"step": 6575
},
{
"epoch": 16.42,
"learning_rate": 3.0438947368421055e-06,
"loss": 0.0002,
"step": 6600
},
{
"epoch": 16.48,
"learning_rate": 3.0215263157894737e-06,
"loss": 0.0002,
"step": 6625
},
{
"epoch": 16.54,
"learning_rate": 2.999157894736842e-06,
"loss": 0.0004,
"step": 6650
},
{
"epoch": 16.6,
"learning_rate": 2.9767894736842106e-06,
"loss": 0.0002,
"step": 6675
},
{
"epoch": 16.67,
"learning_rate": 2.954421052631579e-06,
"loss": 0.0002,
"step": 6700
},
{
"epoch": 16.73,
"learning_rate": 2.932052631578947e-06,
"loss": 0.0002,
"step": 6725
},
{
"epoch": 16.79,
"learning_rate": 2.9096842105263157e-06,
"loss": 0.0002,
"step": 6750
},
{
"epoch": 16.85,
"learning_rate": 2.8873157894736844e-06,
"loss": 0.0005,
"step": 6775
},
{
"epoch": 16.92,
"learning_rate": 2.8649473684210526e-06,
"loss": 0.0002,
"step": 6800
},
{
"epoch": 16.98,
"learning_rate": 2.842578947368421e-06,
"loss": 0.0003,
"step": 6825
},
{
"epoch": 17.04,
"learning_rate": 2.8202105263157895e-06,
"loss": 0.0002,
"step": 6850
},
{
"epoch": 17.1,
"learning_rate": 2.7978421052631578e-06,
"loss": 0.0002,
"step": 6875
},
{
"epoch": 17.16,
"learning_rate": 2.775473684210526e-06,
"loss": 0.0002,
"step": 6900
},
{
"epoch": 17.23,
"learning_rate": 2.753105263157895e-06,
"loss": 0.0002,
"step": 6925
},
{
"epoch": 17.29,
"learning_rate": 2.7307368421052633e-06,
"loss": 0.0002,
"step": 6950
},
{
"epoch": 17.35,
"learning_rate": 2.7083684210526315e-06,
"loss": 0.0002,
"step": 6975
},
{
"epoch": 17.41,
"learning_rate": 2.686e-06,
"loss": 0.0002,
"step": 7000
},
{
"epoch": 17.41,
"eval_loss": 0.49173247814178467,
"eval_runtime": 1108.1321,
"eval_samples_per_second": 1.531,
"eval_steps_per_second": 0.191,
"eval_wer": 19.595096582466567,
"step": 7000
},
{
"epoch": 17.48,
"learning_rate": 2.6636315789473684e-06,
"loss": 0.0002,
"step": 7025
},
{
"epoch": 17.54,
"learning_rate": 2.6412631578947367e-06,
"loss": 0.0002,
"step": 7050
},
{
"epoch": 17.6,
"learning_rate": 2.618894736842105e-06,
"loss": 0.0002,
"step": 7075
},
{
"epoch": 17.66,
"learning_rate": 2.596526315789474e-06,
"loss": 0.0002,
"step": 7100
},
{
"epoch": 17.72,
"learning_rate": 2.5741578947368422e-06,
"loss": 0.0002,
"step": 7125
},
{
"epoch": 17.79,
"learning_rate": 2.5517894736842105e-06,
"loss": 0.0002,
"step": 7150
},
{
"epoch": 17.85,
"learning_rate": 2.529421052631579e-06,
"loss": 0.0002,
"step": 7175
},
{
"epoch": 17.91,
"learning_rate": 2.5070526315789473e-06,
"loss": 0.0003,
"step": 7200
},
{
"epoch": 17.97,
"learning_rate": 2.4846842105263156e-06,
"loss": 0.0002,
"step": 7225
},
{
"epoch": 18.03,
"learning_rate": 2.462315789473684e-06,
"loss": 0.0002,
"step": 7250
},
{
"epoch": 18.1,
"learning_rate": 2.439947368421053e-06,
"loss": 0.0002,
"step": 7275
},
{
"epoch": 18.16,
"learning_rate": 2.417578947368421e-06,
"loss": 0.0002,
"step": 7300
},
{
"epoch": 18.22,
"learning_rate": 2.3952105263157894e-06,
"loss": 0.0002,
"step": 7325
},
{
"epoch": 18.28,
"learning_rate": 2.372842105263158e-06,
"loss": 0.0002,
"step": 7350
},
{
"epoch": 18.35,
"learning_rate": 2.3504736842105263e-06,
"loss": 0.0002,
"step": 7375
},
{
"epoch": 18.41,
"learning_rate": 2.3281052631578945e-06,
"loss": 0.0002,
"step": 7400
},
{
"epoch": 18.47,
"learning_rate": 2.3057368421052627e-06,
"loss": 0.0002,
"step": 7425
},
{
"epoch": 18.53,
"learning_rate": 2.283368421052632e-06,
"loss": 0.0002,
"step": 7450
},
{
"epoch": 18.59,
"learning_rate": 2.261e-06,
"loss": 0.0002,
"step": 7475
},
{
"epoch": 18.66,
"learning_rate": 2.2386315789473683e-06,
"loss": 0.0002,
"step": 7500
},
{
"epoch": 18.72,
"learning_rate": 2.216263157894737e-06,
"loss": 0.0001,
"step": 7525
},
{
"epoch": 18.78,
"learning_rate": 2.193894736842105e-06,
"loss": 0.0002,
"step": 7550
},
{
"epoch": 18.84,
"learning_rate": 2.1715263157894734e-06,
"loss": 0.0002,
"step": 7575
},
{
"epoch": 18.91,
"learning_rate": 2.149157894736842e-06,
"loss": 0.0001,
"step": 7600
},
{
"epoch": 18.97,
"learning_rate": 2.1267894736842107e-06,
"loss": 0.0001,
"step": 7625
},
{
"epoch": 19.03,
"learning_rate": 2.104421052631579e-06,
"loss": 0.0001,
"step": 7650
},
{
"epoch": 19.09,
"learning_rate": 2.082052631578947e-06,
"loss": 0.0001,
"step": 7675
},
{
"epoch": 19.15,
"learning_rate": 2.059684210526316e-06,
"loss": 0.0001,
"step": 7700
},
{
"epoch": 19.22,
"learning_rate": 2.037315789473684e-06,
"loss": 0.0001,
"step": 7725
},
{
"epoch": 19.28,
"learning_rate": 2.0149473684210527e-06,
"loss": 0.0001,
"step": 7750
},
{
"epoch": 19.34,
"learning_rate": 1.992578947368421e-06,
"loss": 0.0001,
"step": 7775
},
{
"epoch": 19.4,
"learning_rate": 1.970210526315789e-06,
"loss": 0.0001,
"step": 7800
},
{
"epoch": 19.47,
"learning_rate": 1.947842105263158e-06,
"loss": 0.0001,
"step": 7825
},
{
"epoch": 19.53,
"learning_rate": 1.9254736842105265e-06,
"loss": 0.0001,
"step": 7850
},
{
"epoch": 19.59,
"learning_rate": 1.9031052631578947e-06,
"loss": 0.0001,
"step": 7875
},
{
"epoch": 19.65,
"learning_rate": 1.880736842105263e-06,
"loss": 0.0001,
"step": 7900
},
{
"epoch": 19.71,
"learning_rate": 1.8583684210526316e-06,
"loss": 0.0001,
"step": 7925
},
{
"epoch": 19.78,
"learning_rate": 1.8359999999999999e-06,
"loss": 0.0001,
"step": 7950
},
{
"epoch": 19.84,
"learning_rate": 1.8136315789473683e-06,
"loss": 0.0001,
"step": 7975
},
{
"epoch": 19.9,
"learning_rate": 1.791263157894737e-06,
"loss": 0.0001,
"step": 8000
},
{
"epoch": 19.9,
"eval_loss": 0.5050469040870667,
"eval_runtime": 1102.3823,
"eval_samples_per_second": 1.538,
"eval_steps_per_second": 0.192,
"eval_wer": 19.62295690936107,
"step": 8000
},
{
"epoch": 19.96,
"learning_rate": 1.7688947368421052e-06,
"loss": 0.0001,
"step": 8025
},
{
"epoch": 20.02,
"learning_rate": 1.7465263157894737e-06,
"loss": 0.0001,
"step": 8050
},
{
"epoch": 20.09,
"learning_rate": 1.724157894736842e-06,
"loss": 0.0001,
"step": 8075
},
{
"epoch": 20.15,
"learning_rate": 1.7017894736842105e-06,
"loss": 0.0001,
"step": 8100
},
{
"epoch": 20.21,
"learning_rate": 1.679421052631579e-06,
"loss": 0.0001,
"step": 8125
},
{
"epoch": 20.27,
"learning_rate": 1.6570526315789472e-06,
"loss": 0.0001,
"step": 8150
},
{
"epoch": 20.34,
"learning_rate": 1.6346842105263159e-06,
"loss": 0.0001,
"step": 8175
},
{
"epoch": 20.4,
"learning_rate": 1.6123157894736841e-06,
"loss": 0.0001,
"step": 8200
},
{
"epoch": 20.46,
"learning_rate": 1.5899473684210526e-06,
"loss": 0.0001,
"step": 8225
},
{
"epoch": 20.52,
"learning_rate": 1.5675789473684212e-06,
"loss": 0.0001,
"step": 8250
},
{
"epoch": 20.58,
"learning_rate": 1.5452105263157895e-06,
"loss": 0.0001,
"step": 8275
},
{
"epoch": 20.65,
"learning_rate": 1.522842105263158e-06,
"loss": 0.0001,
"step": 8300
},
{
"epoch": 20.71,
"learning_rate": 1.5004736842105263e-06,
"loss": 0.0001,
"step": 8325
},
{
"epoch": 20.77,
"learning_rate": 1.4781052631578948e-06,
"loss": 0.0001,
"step": 8350
},
{
"epoch": 20.83,
"learning_rate": 1.455736842105263e-06,
"loss": 0.0001,
"step": 8375
},
{
"epoch": 20.9,
"learning_rate": 1.4333684210526315e-06,
"loss": 0.0001,
"step": 8400
},
{
"epoch": 20.96,
"learning_rate": 1.4110000000000001e-06,
"loss": 0.0001,
"step": 8425
},
{
"epoch": 21.02,
"learning_rate": 1.3886315789473684e-06,
"loss": 0.0001,
"step": 8450
},
{
"epoch": 21.08,
"learning_rate": 1.3662631578947368e-06,
"loss": 0.0001,
"step": 8475
},
{
"epoch": 21.14,
"learning_rate": 1.3438947368421053e-06,
"loss": 0.0001,
"step": 8500
},
{
"epoch": 21.21,
"learning_rate": 1.3215263157894737e-06,
"loss": 0.0001,
"step": 8525
},
{
"epoch": 21.27,
"learning_rate": 1.2991578947368421e-06,
"loss": 0.0001,
"step": 8550
},
{
"epoch": 21.33,
"learning_rate": 1.2767894736842104e-06,
"loss": 0.0001,
"step": 8575
},
{
"epoch": 21.39,
"learning_rate": 1.254421052631579e-06,
"loss": 0.0001,
"step": 8600
},
{
"epoch": 21.46,
"learning_rate": 1.2320526315789473e-06,
"loss": 0.0001,
"step": 8625
},
{
"epoch": 21.52,
"learning_rate": 1.2096842105263157e-06,
"loss": 0.0001,
"step": 8650
},
{
"epoch": 21.58,
"learning_rate": 1.1873157894736844e-06,
"loss": 0.0001,
"step": 8675
},
{
"epoch": 21.64,
"learning_rate": 1.1649473684210526e-06,
"loss": 0.0001,
"step": 8700
},
{
"epoch": 21.7,
"learning_rate": 1.142578947368421e-06,
"loss": 0.0001,
"step": 8725
},
{
"epoch": 21.77,
"learning_rate": 1.1202105263157895e-06,
"loss": 0.0001,
"step": 8750
},
{
"epoch": 21.83,
"learning_rate": 1.097842105263158e-06,
"loss": 0.0001,
"step": 8775
},
{
"epoch": 21.89,
"learning_rate": 1.0754736842105262e-06,
"loss": 0.0001,
"step": 8800
},
{
"epoch": 21.95,
"learning_rate": 1.0531052631578948e-06,
"loss": 0.0001,
"step": 8825
},
{
"epoch": 22.01,
"learning_rate": 1.030736842105263e-06,
"loss": 0.0001,
"step": 8850
},
{
"epoch": 22.08,
"learning_rate": 1.0083684210526315e-06,
"loss": 0.0001,
"step": 8875
},
{
"epoch": 22.14,
"learning_rate": 9.86e-07,
"loss": 0.0001,
"step": 8900
},
{
"epoch": 22.2,
"learning_rate": 9.636315789473684e-07,
"loss": 0.0001,
"step": 8925
},
{
"epoch": 22.26,
"learning_rate": 9.412631578947369e-07,
"loss": 0.0001,
"step": 8950
},
{
"epoch": 22.33,
"learning_rate": 9.188947368421052e-07,
"loss": 0.0001,
"step": 8975
},
{
"epoch": 22.39,
"learning_rate": 8.965263157894736e-07,
"loss": 0.0001,
"step": 9000
},
{
"epoch": 22.39,
"eval_loss": 0.5145551562309265,
"eval_runtime": 1109.8104,
"eval_samples_per_second": 1.528,
"eval_steps_per_second": 0.191,
"eval_wer": 19.567236255572066,
"step": 9000
},
{
"epoch": 22.45,
"learning_rate": 8.741578947368422e-07,
"loss": 0.0001,
"step": 9025
},
{
"epoch": 22.51,
"learning_rate": 8.517894736842105e-07,
"loss": 0.0001,
"step": 9050
},
{
"epoch": 22.57,
"learning_rate": 8.29421052631579e-07,
"loss": 0.0001,
"step": 9075
},
{
"epoch": 22.64,
"learning_rate": 8.070526315789473e-07,
"loss": 0.0001,
"step": 9100
},
{
"epoch": 22.7,
"learning_rate": 7.846842105263158e-07,
"loss": 0.0001,
"step": 9125
},
{
"epoch": 22.76,
"learning_rate": 7.623157894736841e-07,
"loss": 0.0001,
"step": 9150
},
{
"epoch": 22.82,
"learning_rate": 7.399473684210527e-07,
"loss": 0.0001,
"step": 9175
},
{
"epoch": 22.89,
"learning_rate": 7.175789473684211e-07,
"loss": 0.0001,
"step": 9200
},
{
"epoch": 22.95,
"learning_rate": 6.952105263157894e-07,
"loss": 0.0001,
"step": 9225
},
{
"epoch": 23.01,
"learning_rate": 6.728421052631579e-07,
"loss": 0.0001,
"step": 9250
},
{
"epoch": 23.07,
"learning_rate": 6.504736842105262e-07,
"loss": 0.0001,
"step": 9275
},
{
"epoch": 23.13,
"learning_rate": 6.281052631578948e-07,
"loss": 0.0001,
"step": 9300
},
{
"epoch": 23.2,
"learning_rate": 6.057368421052632e-07,
"loss": 0.0001,
"step": 9325
},
{
"epoch": 23.26,
"learning_rate": 5.833684210526316e-07,
"loss": 0.0001,
"step": 9350
},
{
"epoch": 23.32,
"learning_rate": 5.61e-07,
"loss": 0.0001,
"step": 9375
},
{
"epoch": 23.38,
"learning_rate": 5.386315789473683e-07,
"loss": 0.0001,
"step": 9400
},
{
"epoch": 23.45,
"learning_rate": 5.162631578947368e-07,
"loss": 0.0001,
"step": 9425
},
{
"epoch": 23.51,
"learning_rate": 4.938947368421052e-07,
"loss": 0.0001,
"step": 9450
},
{
"epoch": 23.57,
"learning_rate": 4.715263157894737e-07,
"loss": 0.0001,
"step": 9475
},
{
"epoch": 23.63,
"learning_rate": 4.491578947368421e-07,
"loss": 0.0001,
"step": 9500
},
{
"epoch": 23.69,
"learning_rate": 4.267894736842105e-07,
"loss": 0.0001,
"step": 9525
},
{
"epoch": 23.76,
"learning_rate": 4.0442105263157897e-07,
"loss": 0.0001,
"step": 9550
},
{
"epoch": 23.82,
"learning_rate": 3.8205263157894736e-07,
"loss": 0.0001,
"step": 9575
},
{
"epoch": 23.88,
"learning_rate": 3.5968421052631575e-07,
"loss": 0.0001,
"step": 9600
},
{
"epoch": 23.94,
"learning_rate": 3.373157894736842e-07,
"loss": 0.0001,
"step": 9625
},
{
"epoch": 24.0,
"learning_rate": 3.1494736842105265e-07,
"loss": 0.0001,
"step": 9650
},
{
"epoch": 24.07,
"learning_rate": 2.925789473684211e-07,
"loss": 0.0001,
"step": 9675
},
{
"epoch": 24.13,
"learning_rate": 2.702105263157895e-07,
"loss": 0.0001,
"step": 9700
},
{
"epoch": 24.19,
"learning_rate": 2.478421052631579e-07,
"loss": 0.0001,
"step": 9725
},
{
"epoch": 24.25,
"learning_rate": 2.254736842105263e-07,
"loss": 0.0001,
"step": 9750
},
{
"epoch": 24.32,
"learning_rate": 2.0310526315789474e-07,
"loss": 0.0001,
"step": 9775
},
{
"epoch": 24.38,
"learning_rate": 1.8073684210526316e-07,
"loss": 0.0001,
"step": 9800
},
{
"epoch": 24.44,
"learning_rate": 1.5836842105263158e-07,
"loss": 0.0001,
"step": 9825
},
{
"epoch": 24.5,
"learning_rate": 1.36e-07,
"loss": 0.0001,
"step": 9850
},
{
"epoch": 24.56,
"learning_rate": 1.1363157894736843e-07,
"loss": 0.0001,
"step": 9875
},
{
"epoch": 24.63,
"learning_rate": 9.126315789473684e-08,
"loss": 0.0001,
"step": 9900
},
{
"epoch": 24.69,
"learning_rate": 6.889473684210526e-08,
"loss": 0.0001,
"step": 9925
},
{
"epoch": 24.75,
"learning_rate": 4.6526315789473684e-08,
"loss": 0.0001,
"step": 9950
},
{
"epoch": 24.81,
"learning_rate": 2.4157894736842107e-08,
"loss": 0.0001,
"step": 9975
},
{
"epoch": 24.88,
"learning_rate": 1.7894736842105264e-09,
"loss": 0.0001,
"step": 10000
},
{
"epoch": 24.88,
"eval_loss": 0.5186008214950562,
"eval_runtime": 1111.2474,
"eval_samples_per_second": 1.526,
"eval_steps_per_second": 0.191,
"eval_wer": 19.48365527488856,
"step": 10000
},
{
"epoch": 24.88,
"step": 10000,
"total_flos": 9.230577180475392e+19,
"train_loss": 0.02654122557353694,
"train_runtime": 76288.9034,
"train_samples_per_second": 4.195,
"train_steps_per_second": 0.131
}
],
"max_steps": 10000,
"num_train_epochs": 25,
"total_flos": 9.230577180475392e+19,
"trial_name": null,
"trial_params": null
}