agnesluhtaru's picture
Model save
13067d9
raw
history blame
51 kB
{
"best_metric": 15.911884203192397,
"best_model_checkpoint": "models/whisper-medium-et-ERR2020/checkpoint-10000",
"epoch": 1.232,
"global_step": 10000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.0,
"learning_rate": 4.6000000000000004e-07,
"loss": 1.538,
"step": 25
},
{
"epoch": 0.01,
"learning_rate": 9.600000000000001e-07,
"loss": 1.1539,
"step": 50
},
{
"epoch": 0.01,
"learning_rate": 1.46e-06,
"loss": 0.904,
"step": 75
},
{
"epoch": 0.01,
"learning_rate": 1.9600000000000003e-06,
"loss": 0.8887,
"step": 100
},
{
"epoch": 0.01,
"learning_rate": 2.46e-06,
"loss": 0.9294,
"step": 125
},
{
"epoch": 0.01,
"learning_rate": 2.96e-06,
"loss": 0.8664,
"step": 150
},
{
"epoch": 0.02,
"learning_rate": 3.46e-06,
"loss": 0.791,
"step": 175
},
{
"epoch": 0.02,
"learning_rate": 3.96e-06,
"loss": 0.7603,
"step": 200
},
{
"epoch": 0.02,
"learning_rate": 4.4600000000000005e-06,
"loss": 0.6953,
"step": 225
},
{
"epoch": 0.03,
"learning_rate": 4.960000000000001e-06,
"loss": 0.6213,
"step": 250
},
{
"epoch": 0.03,
"learning_rate": 5.460000000000001e-06,
"loss": 0.6298,
"step": 275
},
{
"epoch": 0.03,
"learning_rate": 5.9600000000000005e-06,
"loss": 0.5772,
"step": 300
},
{
"epoch": 0.03,
"learning_rate": 6.460000000000001e-06,
"loss": 0.5241,
"step": 325
},
{
"epoch": 0.04,
"learning_rate": 6.96e-06,
"loss": 0.5019,
"step": 350
},
{
"epoch": 0.04,
"learning_rate": 7.4600000000000006e-06,
"loss": 0.4879,
"step": 375
},
{
"epoch": 0.04,
"learning_rate": 7.960000000000002e-06,
"loss": 0.4563,
"step": 400
},
{
"epoch": 0.04,
"learning_rate": 8.46e-06,
"loss": 0.4498,
"step": 425
},
{
"epoch": 0.04,
"learning_rate": 8.96e-06,
"loss": 0.4136,
"step": 450
},
{
"epoch": 0.05,
"learning_rate": 9.460000000000001e-06,
"loss": 0.4136,
"step": 475
},
{
"epoch": 0.05,
"learning_rate": 9.960000000000001e-06,
"loss": 0.3882,
"step": 500
},
{
"epoch": 0.05,
"learning_rate": 9.975789473684211e-06,
"loss": 0.3335,
"step": 525
},
{
"epoch": 0.06,
"learning_rate": 9.949473684210526e-06,
"loss": 0.3474,
"step": 550
},
{
"epoch": 0.06,
"learning_rate": 9.923157894736844e-06,
"loss": 0.333,
"step": 575
},
{
"epoch": 0.06,
"learning_rate": 9.89684210526316e-06,
"loss": 0.3255,
"step": 600
},
{
"epoch": 0.06,
"learning_rate": 9.870526315789474e-06,
"loss": 0.2756,
"step": 625
},
{
"epoch": 0.07,
"learning_rate": 9.84421052631579e-06,
"loss": 0.281,
"step": 650
},
{
"epoch": 0.07,
"learning_rate": 9.817894736842106e-06,
"loss": 0.2666,
"step": 675
},
{
"epoch": 0.07,
"learning_rate": 9.791578947368422e-06,
"loss": 0.2592,
"step": 700
},
{
"epoch": 0.07,
"learning_rate": 9.765263157894737e-06,
"loss": 0.2504,
"step": 725
},
{
"epoch": 0.07,
"learning_rate": 9.738947368421054e-06,
"loss": 0.2398,
"step": 750
},
{
"epoch": 0.08,
"learning_rate": 9.712631578947369e-06,
"loss": 0.2283,
"step": 775
},
{
"epoch": 0.08,
"learning_rate": 9.686315789473684e-06,
"loss": 0.2377,
"step": 800
},
{
"epoch": 0.08,
"learning_rate": 9.66e-06,
"loss": 0.2262,
"step": 825
},
{
"epoch": 0.09,
"learning_rate": 9.633684210526316e-06,
"loss": 0.1998,
"step": 850
},
{
"epoch": 0.09,
"learning_rate": 9.607368421052632e-06,
"loss": 0.2151,
"step": 875
},
{
"epoch": 0.09,
"learning_rate": 9.581052631578947e-06,
"loss": 0.2035,
"step": 900
},
{
"epoch": 0.09,
"learning_rate": 9.554736842105264e-06,
"loss": 0.189,
"step": 925
},
{
"epoch": 0.1,
"learning_rate": 9.52842105263158e-06,
"loss": 0.1786,
"step": 950
},
{
"epoch": 0.1,
"learning_rate": 9.502105263157896e-06,
"loss": 0.1877,
"step": 975
},
{
"epoch": 0.1,
"learning_rate": 9.475789473684212e-06,
"loss": 0.1828,
"step": 1000
},
{
"epoch": 0.1,
"eval_loss": 0.3546636998653412,
"eval_runtime": 11312.8233,
"eval_samples_per_second": 0.699,
"eval_steps_per_second": 0.044,
"eval_wer": 20.882880202910577,
"step": 1000
},
{
"epoch": 0.1,
"learning_rate": 9.449473684210527e-06,
"loss": 0.1761,
"step": 1025
},
{
"epoch": 0.1,
"learning_rate": 9.423157894736842e-06,
"loss": 0.1627,
"step": 1050
},
{
"epoch": 0.11,
"learning_rate": 9.396842105263159e-06,
"loss": 0.1522,
"step": 1075
},
{
"epoch": 0.11,
"learning_rate": 9.370526315789474e-06,
"loss": 0.1618,
"step": 1100
},
{
"epoch": 0.11,
"learning_rate": 9.34421052631579e-06,
"loss": 0.1485,
"step": 1125
},
{
"epoch": 0.12,
"learning_rate": 9.317894736842105e-06,
"loss": 0.1703,
"step": 1150
},
{
"epoch": 0.12,
"learning_rate": 9.291578947368422e-06,
"loss": 0.148,
"step": 1175
},
{
"epoch": 0.12,
"learning_rate": 9.265263157894737e-06,
"loss": 0.152,
"step": 1200
},
{
"epoch": 0.12,
"learning_rate": 9.238947368421052e-06,
"loss": 0.1459,
"step": 1225
},
{
"epoch": 0.12,
"learning_rate": 9.21263157894737e-06,
"loss": 0.1618,
"step": 1250
},
{
"epoch": 0.13,
"learning_rate": 9.186315789473685e-06,
"loss": 0.1345,
"step": 1275
},
{
"epoch": 0.13,
"learning_rate": 9.16e-06,
"loss": 0.1467,
"step": 1300
},
{
"epoch": 0.13,
"learning_rate": 9.133684210526317e-06,
"loss": 0.1281,
"step": 1325
},
{
"epoch": 0.14,
"learning_rate": 9.107368421052632e-06,
"loss": 0.137,
"step": 1350
},
{
"epoch": 0.14,
"learning_rate": 9.081052631578949e-06,
"loss": 0.1389,
"step": 1375
},
{
"epoch": 0.14,
"learning_rate": 9.054736842105264e-06,
"loss": 0.1306,
"step": 1400
},
{
"epoch": 0.14,
"learning_rate": 9.02842105263158e-06,
"loss": 0.1378,
"step": 1425
},
{
"epoch": 0.14,
"learning_rate": 9.002105263157895e-06,
"loss": 0.1246,
"step": 1450
},
{
"epoch": 0.15,
"learning_rate": 8.97578947368421e-06,
"loss": 0.1325,
"step": 1475
},
{
"epoch": 0.15,
"learning_rate": 8.949473684210527e-06,
"loss": 0.1301,
"step": 1500
},
{
"epoch": 0.15,
"learning_rate": 8.923157894736842e-06,
"loss": 0.1163,
"step": 1525
},
{
"epoch": 0.15,
"learning_rate": 8.896842105263159e-06,
"loss": 0.1233,
"step": 1550
},
{
"epoch": 0.16,
"learning_rate": 8.870526315789474e-06,
"loss": 0.1312,
"step": 1575
},
{
"epoch": 0.16,
"learning_rate": 8.84421052631579e-06,
"loss": 0.115,
"step": 1600
},
{
"epoch": 0.16,
"learning_rate": 8.817894736842107e-06,
"loss": 0.1176,
"step": 1625
},
{
"epoch": 0.17,
"learning_rate": 8.791578947368422e-06,
"loss": 0.109,
"step": 1650
},
{
"epoch": 0.17,
"learning_rate": 8.765263157894739e-06,
"loss": 0.1184,
"step": 1675
},
{
"epoch": 0.17,
"learning_rate": 8.738947368421053e-06,
"loss": 0.1055,
"step": 1700
},
{
"epoch": 0.17,
"learning_rate": 8.712631578947368e-06,
"loss": 0.0995,
"step": 1725
},
{
"epoch": 0.17,
"learning_rate": 8.686315789473685e-06,
"loss": 0.1171,
"step": 1750
},
{
"epoch": 0.18,
"learning_rate": 8.66e-06,
"loss": 0.1222,
"step": 1775
},
{
"epoch": 0.18,
"learning_rate": 8.633684210526317e-06,
"loss": 0.1066,
"step": 1800
},
{
"epoch": 0.18,
"learning_rate": 8.607368421052632e-06,
"loss": 0.1074,
"step": 1825
},
{
"epoch": 0.18,
"learning_rate": 8.581052631578948e-06,
"loss": 0.1064,
"step": 1850
},
{
"epoch": 0.19,
"learning_rate": 8.554736842105263e-06,
"loss": 0.1118,
"step": 1875
},
{
"epoch": 0.19,
"learning_rate": 8.528421052631578e-06,
"loss": 0.1008,
"step": 1900
},
{
"epoch": 0.19,
"learning_rate": 8.502105263157897e-06,
"loss": 0.1069,
"step": 1925
},
{
"epoch": 0.2,
"learning_rate": 8.475789473684212e-06,
"loss": 0.1126,
"step": 1950
},
{
"epoch": 0.2,
"learning_rate": 8.449473684210527e-06,
"loss": 0.1061,
"step": 1975
},
{
"epoch": 0.2,
"learning_rate": 8.423157894736843e-06,
"loss": 0.09,
"step": 2000
},
{
"epoch": 0.2,
"eval_loss": 0.3476346433162689,
"eval_runtime": 11632.5005,
"eval_samples_per_second": 0.68,
"eval_steps_per_second": 0.043,
"eval_wer": 19.00955840333174,
"step": 2000
},
{
"epoch": 0.2,
"learning_rate": 8.396842105263158e-06,
"loss": 0.1139,
"step": 2025
},
{
"epoch": 0.2,
"learning_rate": 8.370526315789475e-06,
"loss": 0.0995,
"step": 2050
},
{
"epoch": 0.21,
"learning_rate": 8.34421052631579e-06,
"loss": 0.1025,
"step": 2075
},
{
"epoch": 0.21,
"learning_rate": 8.317894736842107e-06,
"loss": 0.099,
"step": 2100
},
{
"epoch": 0.21,
"learning_rate": 8.291578947368422e-06,
"loss": 0.0932,
"step": 2125
},
{
"epoch": 0.21,
"learning_rate": 8.265263157894737e-06,
"loss": 0.0998,
"step": 2150
},
{
"epoch": 0.22,
"learning_rate": 8.238947368421053e-06,
"loss": 0.0948,
"step": 2175
},
{
"epoch": 0.22,
"learning_rate": 8.212631578947368e-06,
"loss": 0.0976,
"step": 2200
},
{
"epoch": 0.22,
"learning_rate": 8.186315789473685e-06,
"loss": 0.0936,
"step": 2225
},
{
"epoch": 0.23,
"learning_rate": 8.16e-06,
"loss": 0.0906,
"step": 2250
},
{
"epoch": 0.23,
"learning_rate": 8.133684210526316e-06,
"loss": 0.0918,
"step": 2275
},
{
"epoch": 0.23,
"learning_rate": 8.107368421052633e-06,
"loss": 0.0974,
"step": 2300
},
{
"epoch": 0.23,
"learning_rate": 8.081052631578948e-06,
"loss": 0.1078,
"step": 2325
},
{
"epoch": 0.23,
"learning_rate": 8.054736842105265e-06,
"loss": 0.0928,
"step": 2350
},
{
"epoch": 0.24,
"learning_rate": 8.02842105263158e-06,
"loss": 0.1048,
"step": 2375
},
{
"epoch": 0.24,
"learning_rate": 8.002105263157895e-06,
"loss": 0.0869,
"step": 2400
},
{
"epoch": 0.24,
"learning_rate": 7.975789473684211e-06,
"loss": 0.0872,
"step": 2425
},
{
"epoch": 0.24,
"learning_rate": 7.949473684210526e-06,
"loss": 0.104,
"step": 2450
},
{
"epoch": 0.25,
"learning_rate": 7.923157894736843e-06,
"loss": 0.0897,
"step": 2475
},
{
"epoch": 0.25,
"learning_rate": 7.896842105263158e-06,
"loss": 0.0798,
"step": 2500
},
{
"epoch": 0.25,
"learning_rate": 7.870526315789475e-06,
"loss": 0.0949,
"step": 2525
},
{
"epoch": 0.26,
"learning_rate": 7.84421052631579e-06,
"loss": 0.085,
"step": 2550
},
{
"epoch": 0.26,
"learning_rate": 7.817894736842105e-06,
"loss": 0.0841,
"step": 2575
},
{
"epoch": 0.26,
"learning_rate": 7.791578947368423e-06,
"loss": 0.0795,
"step": 2600
},
{
"epoch": 0.26,
"learning_rate": 7.765263157894738e-06,
"loss": 0.0849,
"step": 2625
},
{
"epoch": 0.27,
"learning_rate": 7.738947368421053e-06,
"loss": 0.0878,
"step": 2650
},
{
"epoch": 0.27,
"learning_rate": 7.71263157894737e-06,
"loss": 0.0858,
"step": 2675
},
{
"epoch": 0.27,
"learning_rate": 7.686315789473685e-06,
"loss": 0.0987,
"step": 2700
},
{
"epoch": 0.27,
"learning_rate": 7.660000000000001e-06,
"loss": 0.088,
"step": 2725
},
{
"epoch": 0.28,
"learning_rate": 7.633684210526316e-06,
"loss": 0.0801,
"step": 2750
},
{
"epoch": 0.28,
"learning_rate": 7.607368421052632e-06,
"loss": 0.089,
"step": 2775
},
{
"epoch": 0.28,
"learning_rate": 7.581052631578948e-06,
"loss": 0.0908,
"step": 2800
},
{
"epoch": 0.28,
"learning_rate": 7.554736842105264e-06,
"loss": 0.0686,
"step": 2825
},
{
"epoch": 0.28,
"learning_rate": 7.5284210526315794e-06,
"loss": 0.0873,
"step": 2850
},
{
"epoch": 0.29,
"learning_rate": 7.502105263157895e-06,
"loss": 0.0913,
"step": 2875
},
{
"epoch": 0.29,
"learning_rate": 7.475789473684211e-06,
"loss": 0.0844,
"step": 2900
},
{
"epoch": 0.29,
"learning_rate": 7.449473684210526e-06,
"loss": 0.0931,
"step": 2925
},
{
"epoch": 0.29,
"learning_rate": 7.4231578947368436e-06,
"loss": 0.0758,
"step": 2950
},
{
"epoch": 0.3,
"learning_rate": 7.3968421052631585e-06,
"loss": 0.0772,
"step": 2975
},
{
"epoch": 0.3,
"learning_rate": 7.370526315789474e-06,
"loss": 0.083,
"step": 3000
},
{
"epoch": 0.3,
"eval_loss": 0.3385966122150421,
"eval_runtime": 11712.6787,
"eval_samples_per_second": 0.676,
"eval_steps_per_second": 0.042,
"eval_wer": 18.130435803696543,
"step": 3000
},
{
"epoch": 0.3,
"learning_rate": 7.34421052631579e-06,
"loss": 0.0862,
"step": 3025
},
{
"epoch": 0.3,
"learning_rate": 7.317894736842106e-06,
"loss": 0.0801,
"step": 3050
},
{
"epoch": 0.31,
"learning_rate": 7.291578947368422e-06,
"loss": 0.0798,
"step": 3075
},
{
"epoch": 0.31,
"learning_rate": 7.265263157894738e-06,
"loss": 0.0854,
"step": 3100
},
{
"epoch": 0.31,
"learning_rate": 7.2389473684210534e-06,
"loss": 0.088,
"step": 3125
},
{
"epoch": 0.32,
"learning_rate": 7.212631578947369e-06,
"loss": 0.0875,
"step": 3150
},
{
"epoch": 0.32,
"learning_rate": 7.186315789473684e-06,
"loss": 0.0801,
"step": 3175
},
{
"epoch": 0.32,
"learning_rate": 7.16e-06,
"loss": 0.0828,
"step": 3200
},
{
"epoch": 0.32,
"learning_rate": 7.133684210526316e-06,
"loss": 0.082,
"step": 3225
},
{
"epoch": 0.33,
"learning_rate": 7.107368421052632e-06,
"loss": 0.0881,
"step": 3250
},
{
"epoch": 0.33,
"learning_rate": 7.0810526315789475e-06,
"loss": 0.0789,
"step": 3275
},
{
"epoch": 0.33,
"learning_rate": 7.054736842105264e-06,
"loss": 0.0835,
"step": 3300
},
{
"epoch": 0.33,
"learning_rate": 7.02842105263158e-06,
"loss": 0.0769,
"step": 3325
},
{
"epoch": 0.34,
"learning_rate": 7.002105263157896e-06,
"loss": 0.0923,
"step": 3350
},
{
"epoch": 0.34,
"learning_rate": 6.975789473684212e-06,
"loss": 0.0765,
"step": 3375
},
{
"epoch": 0.34,
"learning_rate": 6.9494736842105275e-06,
"loss": 0.0867,
"step": 3400
},
{
"epoch": 0.34,
"learning_rate": 6.9231578947368424e-06,
"loss": 0.0761,
"step": 3425
},
{
"epoch": 0.34,
"learning_rate": 6.896842105263158e-06,
"loss": 0.0863,
"step": 3450
},
{
"epoch": 0.35,
"learning_rate": 6.870526315789474e-06,
"loss": 0.0722,
"step": 3475
},
{
"epoch": 0.35,
"learning_rate": 6.84421052631579e-06,
"loss": 0.0799,
"step": 3500
},
{
"epoch": 0.35,
"learning_rate": 6.817894736842106e-06,
"loss": 0.0826,
"step": 3525
},
{
"epoch": 0.35,
"learning_rate": 6.792631578947368e-06,
"loss": 0.0774,
"step": 3550
},
{
"epoch": 0.36,
"learning_rate": 6.766315789473685e-06,
"loss": 0.0791,
"step": 3575
},
{
"epoch": 0.36,
"learning_rate": 6.740000000000001e-06,
"loss": 0.0866,
"step": 3600
},
{
"epoch": 0.36,
"learning_rate": 6.7136842105263165e-06,
"loss": 0.0798,
"step": 3625
},
{
"epoch": 0.36,
"learning_rate": 6.687368421052632e-06,
"loss": 0.0775,
"step": 3650
},
{
"epoch": 0.37,
"learning_rate": 6.661052631578948e-06,
"loss": 0.0833,
"step": 3675
},
{
"epoch": 0.37,
"learning_rate": 6.634736842105264e-06,
"loss": 0.0795,
"step": 3700
},
{
"epoch": 0.37,
"learning_rate": 6.60842105263158e-06,
"loss": 0.0849,
"step": 3725
},
{
"epoch": 0.38,
"learning_rate": 6.582105263157896e-06,
"loss": 0.0768,
"step": 3750
},
{
"epoch": 0.38,
"learning_rate": 6.5557894736842106e-06,
"loss": 0.0685,
"step": 3775
},
{
"epoch": 0.38,
"learning_rate": 6.529473684210526e-06,
"loss": 0.082,
"step": 3800
},
{
"epoch": 0.38,
"learning_rate": 6.503157894736842e-06,
"loss": 0.0738,
"step": 3825
},
{
"epoch": 0.39,
"learning_rate": 6.476842105263158e-06,
"loss": 0.0822,
"step": 3850
},
{
"epoch": 0.39,
"learning_rate": 6.450526315789474e-06,
"loss": 0.0796,
"step": 3875
},
{
"epoch": 0.39,
"learning_rate": 6.42421052631579e-06,
"loss": 0.0691,
"step": 3900
},
{
"epoch": 0.39,
"learning_rate": 6.397894736842106e-06,
"loss": 0.0769,
"step": 3925
},
{
"epoch": 0.4,
"learning_rate": 6.371578947368422e-06,
"loss": 0.0787,
"step": 3950
},
{
"epoch": 0.4,
"learning_rate": 6.345263157894738e-06,
"loss": 0.0846,
"step": 3975
},
{
"epoch": 0.4,
"learning_rate": 6.318947368421054e-06,
"loss": 0.0765,
"step": 4000
},
{
"epoch": 0.4,
"eval_loss": 0.3365060091018677,
"eval_runtime": 11429.3476,
"eval_samples_per_second": 0.692,
"eval_steps_per_second": 0.043,
"eval_wer": 17.25914154421838,
"step": 4000
},
{
"epoch": 0.4,
"learning_rate": 6.292631578947369e-06,
"loss": 0.0736,
"step": 4025
},
{
"epoch": 0.41,
"learning_rate": 6.266315789473685e-06,
"loss": 0.0884,
"step": 4050
},
{
"epoch": 0.41,
"learning_rate": 6.24e-06,
"loss": 0.0795,
"step": 4075
},
{
"epoch": 0.41,
"learning_rate": 6.213684210526316e-06,
"loss": 0.0778,
"step": 4100
},
{
"epoch": 0.41,
"learning_rate": 6.187368421052632e-06,
"loss": 0.0757,
"step": 4125
},
{
"epoch": 0.41,
"learning_rate": 6.161052631578948e-06,
"loss": 0.0866,
"step": 4150
},
{
"epoch": 0.42,
"learning_rate": 6.134736842105264e-06,
"loss": 0.0767,
"step": 4175
},
{
"epoch": 0.42,
"learning_rate": 6.108421052631579e-06,
"loss": 0.0725,
"step": 4200
},
{
"epoch": 0.42,
"learning_rate": 6.0821052631578945e-06,
"loss": 0.0757,
"step": 4225
},
{
"epoch": 0.42,
"learning_rate": 6.055789473684212e-06,
"loss": 0.08,
"step": 4250
},
{
"epoch": 0.43,
"learning_rate": 6.029473684210527e-06,
"loss": 0.0723,
"step": 4275
},
{
"epoch": 0.43,
"learning_rate": 6.003157894736843e-06,
"loss": 0.0801,
"step": 4300
},
{
"epoch": 0.43,
"learning_rate": 5.976842105263159e-06,
"loss": 0.0697,
"step": 4325
},
{
"epoch": 0.43,
"learning_rate": 5.950526315789474e-06,
"loss": 0.0897,
"step": 4350
},
{
"epoch": 0.44,
"learning_rate": 5.92421052631579e-06,
"loss": 0.0798,
"step": 4375
},
{
"epoch": 0.44,
"learning_rate": 5.897894736842106e-06,
"loss": 0.0733,
"step": 4400
},
{
"epoch": 0.44,
"learning_rate": 5.871578947368422e-06,
"loss": 0.0941,
"step": 4425
},
{
"epoch": 0.45,
"learning_rate": 5.845263157894737e-06,
"loss": 0.0676,
"step": 4450
},
{
"epoch": 0.45,
"learning_rate": 5.818947368421053e-06,
"loss": 0.0797,
"step": 4475
},
{
"epoch": 0.45,
"learning_rate": 5.7926315789473685e-06,
"loss": 0.0684,
"step": 4500
},
{
"epoch": 0.45,
"learning_rate": 5.766315789473684e-06,
"loss": 0.0832,
"step": 4525
},
{
"epoch": 0.46,
"learning_rate": 5.74e-06,
"loss": 0.0773,
"step": 4550
},
{
"epoch": 0.46,
"learning_rate": 5.713684210526316e-06,
"loss": 0.0697,
"step": 4575
},
{
"epoch": 0.46,
"learning_rate": 5.687368421052633e-06,
"loss": 0.0806,
"step": 4600
},
{
"epoch": 0.46,
"learning_rate": 5.661052631578948e-06,
"loss": 0.0752,
"step": 4625
},
{
"epoch": 0.47,
"learning_rate": 5.634736842105264e-06,
"loss": 0.0678,
"step": 4650
},
{
"epoch": 0.47,
"learning_rate": 5.60842105263158e-06,
"loss": 0.0754,
"step": 4675
},
{
"epoch": 0.47,
"learning_rate": 5.582105263157895e-06,
"loss": 0.0694,
"step": 4700
},
{
"epoch": 0.47,
"learning_rate": 5.555789473684211e-06,
"loss": 0.0731,
"step": 4725
},
{
"epoch": 0.47,
"learning_rate": 5.529473684210527e-06,
"loss": 0.0779,
"step": 4750
},
{
"epoch": 0.48,
"learning_rate": 5.5031578947368425e-06,
"loss": 0.0847,
"step": 4775
},
{
"epoch": 0.48,
"learning_rate": 5.476842105263158e-06,
"loss": 0.0719,
"step": 4800
},
{
"epoch": 0.48,
"learning_rate": 5.450526315789474e-06,
"loss": 0.0697,
"step": 4825
},
{
"epoch": 0.48,
"learning_rate": 5.42421052631579e-06,
"loss": 0.0643,
"step": 4850
},
{
"epoch": 0.49,
"learning_rate": 5.397894736842105e-06,
"loss": 0.0712,
"step": 4875
},
{
"epoch": 0.49,
"learning_rate": 5.371578947368421e-06,
"loss": 0.078,
"step": 4900
},
{
"epoch": 0.49,
"learning_rate": 5.345263157894738e-06,
"loss": 0.0843,
"step": 4925
},
{
"epoch": 0.49,
"learning_rate": 5.318947368421053e-06,
"loss": 0.0792,
"step": 4950
},
{
"epoch": 0.5,
"learning_rate": 5.292631578947369e-06,
"loss": 0.0722,
"step": 4975
},
{
"epoch": 0.5,
"learning_rate": 5.266315789473685e-06,
"loss": 0.0592,
"step": 5000
},
{
"epoch": 0.5,
"eval_loss": 0.3534376323223114,
"eval_runtime": 13205.6363,
"eval_samples_per_second": 0.599,
"eval_steps_per_second": 0.037,
"eval_wer": 19.021300913567295,
"step": 5000
},
{
"epoch": 0.5,
"learning_rate": 5.240000000000001e-06,
"loss": 0.0754,
"step": 5025
},
{
"epoch": 0.51,
"learning_rate": 5.2136842105263165e-06,
"loss": 0.0759,
"step": 5050
},
{
"epoch": 0.51,
"learning_rate": 5.187368421052632e-06,
"loss": 0.0667,
"step": 5075
},
{
"epoch": 0.51,
"learning_rate": 5.161052631578948e-06,
"loss": 0.0649,
"step": 5100
},
{
"epoch": 0.51,
"learning_rate": 5.134736842105263e-06,
"loss": 0.0723,
"step": 5125
},
{
"epoch": 0.52,
"learning_rate": 5.108421052631579e-06,
"loss": 0.0756,
"step": 5150
},
{
"epoch": 0.52,
"learning_rate": 5.082105263157895e-06,
"loss": 0.0787,
"step": 5175
},
{
"epoch": 0.52,
"learning_rate": 5.0557894736842105e-06,
"loss": 0.0752,
"step": 5200
},
{
"epoch": 0.52,
"learning_rate": 5.029473684210526e-06,
"loss": 0.0755,
"step": 5225
},
{
"epoch": 0.53,
"learning_rate": 5.003157894736842e-06,
"loss": 0.0773,
"step": 5250
},
{
"epoch": 0.53,
"learning_rate": 4.976842105263158e-06,
"loss": 0.063,
"step": 5275
},
{
"epoch": 0.53,
"learning_rate": 4.950526315789474e-06,
"loss": 0.0701,
"step": 5300
},
{
"epoch": 0.53,
"learning_rate": 4.92421052631579e-06,
"loss": 0.0673,
"step": 5325
},
{
"epoch": 0.54,
"learning_rate": 4.897894736842106e-06,
"loss": 0.0742,
"step": 5350
},
{
"epoch": 0.54,
"learning_rate": 4.871578947368421e-06,
"loss": 0.0754,
"step": 5375
},
{
"epoch": 0.54,
"learning_rate": 4.845263157894737e-06,
"loss": 0.0878,
"step": 5400
},
{
"epoch": 0.54,
"learning_rate": 4.818947368421053e-06,
"loss": 0.0663,
"step": 5425
},
{
"epoch": 0.55,
"learning_rate": 4.792631578947369e-06,
"loss": 0.0711,
"step": 5450
},
{
"epoch": 0.55,
"learning_rate": 4.7663157894736845e-06,
"loss": 0.0636,
"step": 5475
},
{
"epoch": 0.55,
"learning_rate": 4.74e-06,
"loss": 0.0719,
"step": 5500
},
{
"epoch": 0.55,
"learning_rate": 4.713684210526316e-06,
"loss": 0.0711,
"step": 5525
},
{
"epoch": 0.56,
"learning_rate": 4.687368421052632e-06,
"loss": 0.0656,
"step": 5550
},
{
"epoch": 0.56,
"learning_rate": 4.661052631578948e-06,
"loss": 0.0741,
"step": 5575
},
{
"epoch": 0.56,
"learning_rate": 4.634736842105264e-06,
"loss": 0.0639,
"step": 5600
},
{
"epoch": 0.56,
"learning_rate": 4.6084210526315794e-06,
"loss": 0.0716,
"step": 5625
},
{
"epoch": 0.56,
"learning_rate": 4.582105263157895e-06,
"loss": 0.0735,
"step": 5650
},
{
"epoch": 0.57,
"learning_rate": 4.55578947368421e-06,
"loss": 0.0677,
"step": 5675
},
{
"epoch": 0.57,
"learning_rate": 4.529473684210527e-06,
"loss": 0.0615,
"step": 5700
},
{
"epoch": 0.57,
"learning_rate": 4.503157894736843e-06,
"loss": 0.0692,
"step": 5725
},
{
"epoch": 0.57,
"learning_rate": 4.4768421052631585e-06,
"loss": 0.0764,
"step": 5750
},
{
"epoch": 0.58,
"learning_rate": 4.450526315789474e-06,
"loss": 0.0661,
"step": 5775
},
{
"epoch": 0.58,
"learning_rate": 4.424210526315789e-06,
"loss": 0.0703,
"step": 5800
},
{
"epoch": 0.58,
"learning_rate": 4.397894736842105e-06,
"loss": 0.075,
"step": 5825
},
{
"epoch": 0.58,
"learning_rate": 4.371578947368421e-06,
"loss": 0.0679,
"step": 5850
},
{
"epoch": 0.59,
"learning_rate": 4.345263157894738e-06,
"loss": 0.0685,
"step": 5875
},
{
"epoch": 0.59,
"learning_rate": 4.3189473684210535e-06,
"loss": 0.0673,
"step": 5900
},
{
"epoch": 0.59,
"learning_rate": 4.2926315789473684e-06,
"loss": 0.0702,
"step": 5925
},
{
"epoch": 0.59,
"learning_rate": 4.266315789473684e-06,
"loss": 0.0726,
"step": 5950
},
{
"epoch": 0.6,
"learning_rate": 4.24e-06,
"loss": 0.0595,
"step": 5975
},
{
"epoch": 0.6,
"learning_rate": 4.213684210526316e-06,
"loss": 0.0672,
"step": 6000
},
{
"epoch": 0.6,
"eval_loss": 0.36219528317451477,
"eval_runtime": 13383.7299,
"eval_samples_per_second": 0.591,
"eval_steps_per_second": 0.037,
"eval_wer": 18.426347061632523,
"step": 6000
},
{
"epoch": 0.6,
"learning_rate": 4.1873684210526325e-06,
"loss": 0.0628,
"step": 6025
},
{
"epoch": 0.6,
"learning_rate": 4.1610526315789475e-06,
"loss": 0.061,
"step": 6050
},
{
"epoch": 0.61,
"learning_rate": 4.134736842105263e-06,
"loss": 0.0822,
"step": 6075
},
{
"epoch": 0.61,
"learning_rate": 4.108421052631579e-06,
"loss": 0.0664,
"step": 6100
},
{
"epoch": 0.61,
"learning_rate": 4.082105263157895e-06,
"loss": 0.0766,
"step": 6125
},
{
"epoch": 0.61,
"learning_rate": 4.055789473684211e-06,
"loss": 0.0699,
"step": 6150
},
{
"epoch": 0.62,
"learning_rate": 4.029473684210527e-06,
"loss": 0.0764,
"step": 6175
},
{
"epoch": 0.62,
"learning_rate": 4.0031578947368424e-06,
"loss": 0.0624,
"step": 6200
},
{
"epoch": 0.62,
"learning_rate": 3.976842105263158e-06,
"loss": 0.0728,
"step": 6225
},
{
"epoch": 0.62,
"learning_rate": 3.950526315789474e-06,
"loss": 0.0728,
"step": 6250
},
{
"epoch": 0.63,
"learning_rate": 3.92421052631579e-06,
"loss": 0.0684,
"step": 6275
},
{
"epoch": 0.63,
"learning_rate": 3.897894736842106e-06,
"loss": 0.0573,
"step": 6300
},
{
"epoch": 0.63,
"learning_rate": 3.8715789473684215e-06,
"loss": 0.0644,
"step": 6325
},
{
"epoch": 0.64,
"learning_rate": 3.845263157894737e-06,
"loss": 0.0783,
"step": 6350
},
{
"epoch": 0.64,
"learning_rate": 3.818947368421053e-06,
"loss": 0.0712,
"step": 6375
},
{
"epoch": 0.64,
"learning_rate": 3.793684210526316e-06,
"loss": 0.0711,
"step": 6400
},
{
"epoch": 0.64,
"learning_rate": 3.767368421052632e-06,
"loss": 0.0683,
"step": 6425
},
{
"epoch": 0.65,
"learning_rate": 3.7410526315789473e-06,
"loss": 0.0611,
"step": 6450
},
{
"epoch": 0.65,
"learning_rate": 3.714736842105263e-06,
"loss": 0.0715,
"step": 6475
},
{
"epoch": 0.65,
"learning_rate": 3.6884210526315794e-06,
"loss": 0.0599,
"step": 6500
},
{
"epoch": 0.65,
"learning_rate": 3.662105263157895e-06,
"loss": 0.0738,
"step": 6525
},
{
"epoch": 0.66,
"learning_rate": 3.635789473684211e-06,
"loss": 0.0724,
"step": 6550
},
{
"epoch": 0.66,
"learning_rate": 3.6094736842105264e-06,
"loss": 0.0709,
"step": 6575
},
{
"epoch": 0.66,
"learning_rate": 3.5831578947368422e-06,
"loss": 0.0679,
"step": 6600
},
{
"epoch": 0.66,
"learning_rate": 3.556842105263158e-06,
"loss": 0.0728,
"step": 6625
},
{
"epoch": 0.67,
"learning_rate": 3.5305263157894743e-06,
"loss": 0.065,
"step": 6650
},
{
"epoch": 0.67,
"learning_rate": 3.50421052631579e-06,
"loss": 0.07,
"step": 6675
},
{
"epoch": 0.67,
"learning_rate": 3.4778947368421055e-06,
"loss": 0.0756,
"step": 6700
},
{
"epoch": 0.67,
"learning_rate": 3.4515789473684213e-06,
"loss": 0.0737,
"step": 6725
},
{
"epoch": 0.68,
"learning_rate": 3.425263157894737e-06,
"loss": 0.0612,
"step": 6750
},
{
"epoch": 0.68,
"learning_rate": 3.398947368421053e-06,
"loss": 0.0676,
"step": 6775
},
{
"epoch": 0.68,
"learning_rate": 3.3726315789473683e-06,
"loss": 0.0675,
"step": 6800
},
{
"epoch": 0.68,
"learning_rate": 3.3463157894736846e-06,
"loss": 0.0661,
"step": 6825
},
{
"epoch": 0.69,
"learning_rate": 3.3200000000000004e-06,
"loss": 0.0676,
"step": 6850
},
{
"epoch": 0.69,
"learning_rate": 3.2936842105263162e-06,
"loss": 0.0692,
"step": 6875
},
{
"epoch": 0.69,
"learning_rate": 3.267368421052632e-06,
"loss": 0.0641,
"step": 6900
},
{
"epoch": 0.69,
"learning_rate": 3.2410526315789474e-06,
"loss": 0.0677,
"step": 6925
},
{
"epoch": 0.69,
"learning_rate": 3.2147368421052633e-06,
"loss": 0.0754,
"step": 6950
},
{
"epoch": 0.7,
"learning_rate": 3.188421052631579e-06,
"loss": 0.0691,
"step": 6975
},
{
"epoch": 0.7,
"learning_rate": 3.1621052631578953e-06,
"loss": 0.0629,
"step": 7000
},
{
"epoch": 0.7,
"eval_loss": 0.34873852133750916,
"eval_runtime": 11449.2801,
"eval_samples_per_second": 0.691,
"eval_steps_per_second": 0.043,
"eval_wer": 15.983904932637133,
"step": 7000
},
{
"epoch": 0.7,
"learning_rate": 3.135789473684211e-06,
"loss": 0.0694,
"step": 7025
},
{
"epoch": 0.7,
"learning_rate": 3.1094736842105265e-06,
"loss": 0.0604,
"step": 7050
},
{
"epoch": 0.71,
"learning_rate": 3.0831578947368423e-06,
"loss": 0.0621,
"step": 7075
},
{
"epoch": 0.71,
"learning_rate": 3.056842105263158e-06,
"loss": 0.0598,
"step": 7100
},
{
"epoch": 0.71,
"learning_rate": 3.0305263157894736e-06,
"loss": 0.0732,
"step": 7125
},
{
"epoch": 0.71,
"learning_rate": 3.0042105263157894e-06,
"loss": 0.0649,
"step": 7150
},
{
"epoch": 0.72,
"learning_rate": 2.9778947368421056e-06,
"loss": 0.068,
"step": 7175
},
{
"epoch": 0.72,
"learning_rate": 2.9515789473684214e-06,
"loss": 0.0613,
"step": 7200
},
{
"epoch": 0.72,
"learning_rate": 2.9252631578947373e-06,
"loss": 0.0747,
"step": 7225
},
{
"epoch": 0.72,
"learning_rate": 2.8989473684210526e-06,
"loss": 0.0677,
"step": 7250
},
{
"epoch": 0.73,
"learning_rate": 2.8726315789473685e-06,
"loss": 0.0713,
"step": 7275
},
{
"epoch": 0.73,
"learning_rate": 2.8463157894736843e-06,
"loss": 0.0662,
"step": 7300
},
{
"epoch": 0.73,
"learning_rate": 2.82e-06,
"loss": 0.0666,
"step": 7325
},
{
"epoch": 0.73,
"learning_rate": 2.7936842105263163e-06,
"loss": 0.0589,
"step": 7350
},
{
"epoch": 0.74,
"learning_rate": 2.7673684210526317e-06,
"loss": 0.0667,
"step": 7375
},
{
"epoch": 0.74,
"learning_rate": 2.7410526315789476e-06,
"loss": 0.0655,
"step": 7400
},
{
"epoch": 0.74,
"learning_rate": 2.7147368421052634e-06,
"loss": 0.0678,
"step": 7425
},
{
"epoch": 0.74,
"learning_rate": 2.688421052631579e-06,
"loss": 0.0625,
"step": 7450
},
{
"epoch": 0.75,
"learning_rate": 2.6621052631578946e-06,
"loss": 0.0562,
"step": 7475
},
{
"epoch": 0.75,
"learning_rate": 2.635789473684211e-06,
"loss": 0.0747,
"step": 7500
},
{
"epoch": 0.75,
"learning_rate": 2.6094736842105267e-06,
"loss": 0.0578,
"step": 7525
},
{
"epoch": 0.76,
"learning_rate": 2.5831578947368425e-06,
"loss": 0.0751,
"step": 7550
},
{
"epoch": 0.76,
"learning_rate": 2.5568421052631583e-06,
"loss": 0.0596,
"step": 7575
},
{
"epoch": 0.76,
"learning_rate": 2.5305263157894737e-06,
"loss": 0.0702,
"step": 7600
},
{
"epoch": 0.76,
"learning_rate": 2.5042105263157895e-06,
"loss": 0.0657,
"step": 7625
},
{
"epoch": 0.77,
"learning_rate": 2.4778947368421053e-06,
"loss": 0.0646,
"step": 7650
},
{
"epoch": 0.77,
"learning_rate": 2.451578947368421e-06,
"loss": 0.0625,
"step": 7675
},
{
"epoch": 1.0,
"learning_rate": 2.425263157894737e-06,
"loss": 0.0654,
"step": 7700
},
{
"epoch": 1.0,
"learning_rate": 2.3989473684210528e-06,
"loss": 0.0675,
"step": 7725
},
{
"epoch": 1.01,
"learning_rate": 2.3726315789473686e-06,
"loss": 0.0684,
"step": 7750
},
{
"epoch": 1.01,
"learning_rate": 2.3463157894736844e-06,
"loss": 0.0672,
"step": 7775
},
{
"epoch": 1.01,
"learning_rate": 2.3200000000000002e-06,
"loss": 0.0667,
"step": 7800
},
{
"epoch": 1.01,
"learning_rate": 2.293684210526316e-06,
"loss": 0.0554,
"step": 7825
},
{
"epoch": 1.02,
"learning_rate": 2.267368421052632e-06,
"loss": 0.0583,
"step": 7850
},
{
"epoch": 1.02,
"learning_rate": 2.2410526315789473e-06,
"loss": 0.0592,
"step": 7875
},
{
"epoch": 1.02,
"learning_rate": 2.2147368421052635e-06,
"loss": 0.0637,
"step": 7900
},
{
"epoch": 1.02,
"learning_rate": 2.188421052631579e-06,
"loss": 0.0614,
"step": 7925
},
{
"epoch": 1.03,
"learning_rate": 2.1621052631578947e-06,
"loss": 0.0652,
"step": 7950
},
{
"epoch": 1.03,
"learning_rate": 2.135789473684211e-06,
"loss": 0.0595,
"step": 7975
},
{
"epoch": 1.03,
"learning_rate": 2.1094736842105264e-06,
"loss": 0.0546,
"step": 8000
},
{
"epoch": 1.03,
"eval_loss": 0.3676784038543701,
"eval_runtime": 11276.841,
"eval_samples_per_second": 0.702,
"eval_steps_per_second": 0.044,
"eval_wer": 16.102112869008383,
"step": 8000
},
{
"epoch": 1.03,
"learning_rate": 2.083157894736842e-06,
"loss": 0.063,
"step": 8025
},
{
"epoch": 1.04,
"learning_rate": 2.056842105263158e-06,
"loss": 0.0576,
"step": 8050
},
{
"epoch": 1.04,
"learning_rate": 2.030526315789474e-06,
"loss": 0.0565,
"step": 8075
},
{
"epoch": 1.04,
"learning_rate": 2.0042105263157896e-06,
"loss": 0.0517,
"step": 8100
},
{
"epoch": 1.04,
"learning_rate": 1.9778947368421055e-06,
"loss": 0.0518,
"step": 8125
},
{
"epoch": 1.05,
"learning_rate": 1.9515789473684213e-06,
"loss": 0.0509,
"step": 8150
},
{
"epoch": 1.05,
"learning_rate": 1.925263157894737e-06,
"loss": 0.055,
"step": 8175
},
{
"epoch": 1.05,
"learning_rate": 1.8989473684210527e-06,
"loss": 0.0454,
"step": 8200
},
{
"epoch": 1.05,
"learning_rate": 1.8726315789473687e-06,
"loss": 0.0586,
"step": 8225
},
{
"epoch": 1.06,
"learning_rate": 1.8463157894736843e-06,
"loss": 0.0553,
"step": 8250
},
{
"epoch": 1.06,
"learning_rate": 1.8200000000000002e-06,
"loss": 0.0501,
"step": 8275
},
{
"epoch": 1.06,
"learning_rate": 1.7936842105263158e-06,
"loss": 0.0449,
"step": 8300
},
{
"epoch": 1.06,
"learning_rate": 1.7673684210526318e-06,
"loss": 0.0458,
"step": 8325
},
{
"epoch": 1.07,
"learning_rate": 1.7410526315789474e-06,
"loss": 0.046,
"step": 8350
},
{
"epoch": 1.07,
"learning_rate": 1.7147368421052632e-06,
"loss": 0.0515,
"step": 8375
},
{
"epoch": 1.07,
"learning_rate": 1.6884210526315792e-06,
"loss": 0.0511,
"step": 8400
},
{
"epoch": 1.07,
"learning_rate": 1.6621052631578948e-06,
"loss": 0.0463,
"step": 8425
},
{
"epoch": 1.08,
"learning_rate": 1.6357894736842107e-06,
"loss": 0.0492,
"step": 8450
},
{
"epoch": 1.08,
"learning_rate": 1.6094736842105265e-06,
"loss": 0.0515,
"step": 8475
},
{
"epoch": 1.08,
"learning_rate": 1.5831578947368423e-06,
"loss": 0.0435,
"step": 8500
},
{
"epoch": 1.08,
"learning_rate": 1.556842105263158e-06,
"loss": 0.0473,
"step": 8525
},
{
"epoch": 1.09,
"learning_rate": 1.5305263157894737e-06,
"loss": 0.0488,
"step": 8550
},
{
"epoch": 1.09,
"learning_rate": 1.5042105263157898e-06,
"loss": 0.049,
"step": 8575
},
{
"epoch": 1.09,
"learning_rate": 1.4778947368421054e-06,
"loss": 0.0465,
"step": 8600
},
{
"epoch": 1.09,
"learning_rate": 1.4515789473684212e-06,
"loss": 0.0442,
"step": 8625
},
{
"epoch": 1.1,
"learning_rate": 1.425263157894737e-06,
"loss": 0.0505,
"step": 8650
},
{
"epoch": 1.1,
"learning_rate": 1.3989473684210528e-06,
"loss": 0.038,
"step": 8675
},
{
"epoch": 1.1,
"learning_rate": 1.3726315789473684e-06,
"loss": 0.0457,
"step": 8700
},
{
"epoch": 1.1,
"learning_rate": 1.3463157894736842e-06,
"loss": 0.0441,
"step": 8725
},
{
"epoch": 1.11,
"learning_rate": 1.32e-06,
"loss": 0.0456,
"step": 8750
},
{
"epoch": 1.11,
"learning_rate": 1.2936842105263159e-06,
"loss": 0.0422,
"step": 8775
},
{
"epoch": 1.11,
"learning_rate": 1.2673684210526315e-06,
"loss": 0.0444,
"step": 8800
},
{
"epoch": 1.11,
"learning_rate": 1.2410526315789475e-06,
"loss": 0.0462,
"step": 8825
},
{
"epoch": 1.12,
"learning_rate": 1.2147368421052633e-06,
"loss": 0.046,
"step": 8850
},
{
"epoch": 1.12,
"learning_rate": 1.188421052631579e-06,
"loss": 0.0449,
"step": 8875
},
{
"epoch": 1.12,
"learning_rate": 1.1621052631578948e-06,
"loss": 0.0397,
"step": 8900
},
{
"epoch": 1.12,
"learning_rate": 1.1357894736842106e-06,
"loss": 0.0442,
"step": 8925
},
{
"epoch": 1.13,
"learning_rate": 1.1094736842105264e-06,
"loss": 0.0434,
"step": 8950
},
{
"epoch": 1.13,
"learning_rate": 1.0831578947368422e-06,
"loss": 0.0464,
"step": 8975
},
{
"epoch": 1.13,
"learning_rate": 1.0568421052631578e-06,
"loss": 0.0459,
"step": 9000
},
{
"epoch": 1.13,
"eval_loss": 0.3703891336917877,
"eval_runtime": 12802.7178,
"eval_samples_per_second": 0.618,
"eval_steps_per_second": 0.039,
"eval_wer": 17.907328109221,
"step": 9000
},
{
"epoch": 1.13,
"learning_rate": 1.0305263157894739e-06,
"loss": 0.0455,
"step": 9025
},
{
"epoch": 1.14,
"learning_rate": 1.0042105263157897e-06,
"loss": 0.0456,
"step": 9050
},
{
"epoch": 1.14,
"learning_rate": 9.778947368421053e-07,
"loss": 0.0403,
"step": 9075
},
{
"epoch": 1.14,
"learning_rate": 9.515789473684212e-07,
"loss": 0.0429,
"step": 9100
},
{
"epoch": 1.14,
"learning_rate": 9.252631578947368e-07,
"loss": 0.0418,
"step": 9125
},
{
"epoch": 1.15,
"learning_rate": 8.989473684210527e-07,
"loss": 0.0483,
"step": 9150
},
{
"epoch": 1.15,
"learning_rate": 8.726315789473686e-07,
"loss": 0.0473,
"step": 9175
},
{
"epoch": 1.15,
"learning_rate": 8.463157894736843e-07,
"loss": 0.0409,
"step": 9200
},
{
"epoch": 1.15,
"learning_rate": 8.200000000000001e-07,
"loss": 0.0421,
"step": 9225
},
{
"epoch": 1.16,
"learning_rate": 7.936842105263158e-07,
"loss": 0.0507,
"step": 9250
},
{
"epoch": 1.16,
"learning_rate": 7.673684210526316e-07,
"loss": 0.0361,
"step": 9275
},
{
"epoch": 1.16,
"learning_rate": 7.410526315789475e-07,
"loss": 0.0427,
"step": 9300
},
{
"epoch": 1.16,
"learning_rate": 7.147368421052632e-07,
"loss": 0.0396,
"step": 9325
},
{
"epoch": 1.17,
"learning_rate": 6.884210526315791e-07,
"loss": 0.0383,
"step": 9350
},
{
"epoch": 1.17,
"learning_rate": 6.621052631578948e-07,
"loss": 0.0405,
"step": 9375
},
{
"epoch": 1.17,
"learning_rate": 6.357894736842106e-07,
"loss": 0.0363,
"step": 9400
},
{
"epoch": 1.17,
"learning_rate": 6.094736842105263e-07,
"loss": 0.0426,
"step": 9425
},
{
"epoch": 1.18,
"learning_rate": 5.831578947368421e-07,
"loss": 0.0441,
"step": 9450
},
{
"epoch": 1.18,
"learning_rate": 5.56842105263158e-07,
"loss": 0.0404,
"step": 9475
},
{
"epoch": 1.18,
"learning_rate": 5.305263157894737e-07,
"loss": 0.0447,
"step": 9500
},
{
"epoch": 1.18,
"learning_rate": 5.042105263157895e-07,
"loss": 0.0393,
"step": 9525
},
{
"epoch": 1.19,
"learning_rate": 4.778947368421053e-07,
"loss": 0.0465,
"step": 9550
},
{
"epoch": 1.19,
"learning_rate": 4.5157894736842107e-07,
"loss": 0.0364,
"step": 9575
},
{
"epoch": 1.19,
"learning_rate": 4.2526315789473684e-07,
"loss": 0.0418,
"step": 9600
},
{
"epoch": 1.19,
"learning_rate": 3.9894736842105266e-07,
"loss": 0.0401,
"step": 9625
},
{
"epoch": 1.2,
"learning_rate": 3.726315789473685e-07,
"loss": 0.0374,
"step": 9650
},
{
"epoch": 1.2,
"learning_rate": 3.4631578947368424e-07,
"loss": 0.0387,
"step": 9675
},
{
"epoch": 1.2,
"learning_rate": 3.2e-07,
"loss": 0.0414,
"step": 9700
},
{
"epoch": 1.2,
"learning_rate": 2.936842105263158e-07,
"loss": 0.042,
"step": 9725
},
{
"epoch": 1.21,
"learning_rate": 2.6736842105263164e-07,
"loss": 0.0461,
"step": 9750
},
{
"epoch": 1.21,
"learning_rate": 2.410526315789474e-07,
"loss": 0.0396,
"step": 9775
},
{
"epoch": 1.21,
"learning_rate": 2.1473684210526317e-07,
"loss": 0.0421,
"step": 9800
},
{
"epoch": 1.21,
"learning_rate": 1.8842105263157897e-07,
"loss": 0.0355,
"step": 9825
},
{
"epoch": 1.22,
"learning_rate": 1.6210526315789476e-07,
"loss": 0.0363,
"step": 9850
},
{
"epoch": 1.22,
"learning_rate": 1.3578947368421055e-07,
"loss": 0.0358,
"step": 9875
},
{
"epoch": 1.22,
"learning_rate": 1.0947368421052632e-07,
"loss": 0.0438,
"step": 9900
},
{
"epoch": 1.22,
"learning_rate": 8.315789473684211e-08,
"loss": 0.0367,
"step": 9925
},
{
"epoch": 1.23,
"learning_rate": 5.68421052631579e-08,
"loss": 0.0361,
"step": 9950
},
{
"epoch": 1.23,
"learning_rate": 3.0526315789473686e-08,
"loss": 0.0443,
"step": 9975
},
{
"epoch": 1.23,
"learning_rate": 4.210526315789474e-09,
"loss": 0.0425,
"step": 10000
},
{
"epoch": 1.23,
"eval_loss": 0.3672083616256714,
"eval_runtime": 11298.9209,
"eval_samples_per_second": 0.7,
"eval_steps_per_second": 0.044,
"eval_wer": 15.911884203192397,
"step": 10000
}
],
"max_steps": 10000,
"num_train_epochs": 9223372036854775807,
"total_flos": 3.2658642617204736e+20,
"trial_name": null,
"trial_params": null
}