whisper-small-zh-TW / trainer_state.json
xmzhu's picture
End of training
d13a747
raw
history blame contribute delete
No virus
25.9 kB
{
"best_metric": 42.988741044012286,
"best_model_checkpoint": "./checkpoint-2000",
"epoch": 33.0232,
"global_step": 5000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.01,
"learning_rate": 4.6000000000000004e-07,
"loss": 1.4463,
"step": 25
},
{
"epoch": 0.01,
"learning_rate": 9.600000000000001e-07,
"loss": 0.6808,
"step": 50
},
{
"epoch": 0.01,
"learning_rate": 1.46e-06,
"loss": 0.3478,
"step": 75
},
{
"epoch": 0.02,
"learning_rate": 1.9600000000000003e-06,
"loss": 0.2961,
"step": 100
},
{
"epoch": 0.03,
"learning_rate": 2.46e-06,
"loss": 0.2988,
"step": 125
},
{
"epoch": 1.0,
"learning_rate": 2.96e-06,
"loss": 0.2817,
"step": 150
},
{
"epoch": 1.01,
"learning_rate": 3.46e-06,
"loss": 0.2357,
"step": 175
},
{
"epoch": 1.01,
"learning_rate": 3.96e-06,
"loss": 0.2119,
"step": 200
},
{
"epoch": 1.02,
"learning_rate": 4.4600000000000005e-06,
"loss": 0.1838,
"step": 225
},
{
"epoch": 1.02,
"learning_rate": 4.960000000000001e-06,
"loss": 0.1709,
"step": 250
},
{
"epoch": 1.03,
"learning_rate": 5.460000000000001e-06,
"loss": 0.174,
"step": 275
},
{
"epoch": 2.0,
"learning_rate": 5.9600000000000005e-06,
"loss": 0.1566,
"step": 300
},
{
"epoch": 2.01,
"learning_rate": 6.460000000000001e-06,
"loss": 0.124,
"step": 325
},
{
"epoch": 2.01,
"learning_rate": 6.96e-06,
"loss": 0.1123,
"step": 350
},
{
"epoch": 2.02,
"learning_rate": 7.4600000000000006e-06,
"loss": 0.0932,
"step": 375
},
{
"epoch": 2.02,
"learning_rate": 7.960000000000002e-06,
"loss": 0.0896,
"step": 400
},
{
"epoch": 2.03,
"learning_rate": 8.46e-06,
"loss": 0.0839,
"step": 425
},
{
"epoch": 3.0,
"learning_rate": 8.96e-06,
"loss": 0.0742,
"step": 450
},
{
"epoch": 3.01,
"learning_rate": 9.460000000000001e-06,
"loss": 0.0533,
"step": 475
},
{
"epoch": 3.01,
"learning_rate": 9.960000000000001e-06,
"loss": 0.0454,
"step": 500
},
{
"epoch": 3.02,
"learning_rate": 9.94888888888889e-06,
"loss": 0.0392,
"step": 525
},
{
"epoch": 3.02,
"learning_rate": 9.893333333333334e-06,
"loss": 0.0352,
"step": 550
},
{
"epoch": 3.03,
"learning_rate": 9.837777777777778e-06,
"loss": 0.034,
"step": 575
},
{
"epoch": 4.0,
"learning_rate": 9.782222222222222e-06,
"loss": 0.0283,
"step": 600
},
{
"epoch": 4.01,
"learning_rate": 9.726666666666668e-06,
"loss": 0.0207,
"step": 625
},
{
"epoch": 4.01,
"learning_rate": 9.671111111111112e-06,
"loss": 0.0183,
"step": 650
},
{
"epoch": 4.02,
"learning_rate": 9.615555555555558e-06,
"loss": 0.019,
"step": 675
},
{
"epoch": 4.02,
"learning_rate": 9.56e-06,
"loss": 0.0163,
"step": 700
},
{
"epoch": 4.03,
"learning_rate": 9.504444444444446e-06,
"loss": 0.0163,
"step": 725
},
{
"epoch": 5.0,
"learning_rate": 9.44888888888889e-06,
"loss": 0.0138,
"step": 750
},
{
"epoch": 5.01,
"learning_rate": 9.393333333333334e-06,
"loss": 0.0107,
"step": 775
},
{
"epoch": 5.01,
"learning_rate": 9.33777777777778e-06,
"loss": 0.0095,
"step": 800
},
{
"epoch": 5.02,
"learning_rate": 9.282222222222222e-06,
"loss": 0.0095,
"step": 825
},
{
"epoch": 5.02,
"learning_rate": 9.226666666666668e-06,
"loss": 0.0082,
"step": 850
},
{
"epoch": 5.03,
"learning_rate": 9.171111111111112e-06,
"loss": 0.0073,
"step": 875
},
{
"epoch": 6.0,
"learning_rate": 9.115555555555556e-06,
"loss": 0.0057,
"step": 900
},
{
"epoch": 6.01,
"learning_rate": 9.060000000000001e-06,
"loss": 0.0048,
"step": 925
},
{
"epoch": 6.01,
"learning_rate": 9.004444444444445e-06,
"loss": 0.0068,
"step": 950
},
{
"epoch": 6.02,
"learning_rate": 8.94888888888889e-06,
"loss": 0.0058,
"step": 975
},
{
"epoch": 6.02,
"learning_rate": 8.893333333333333e-06,
"loss": 0.0058,
"step": 1000
},
{
"epoch": 6.02,
"eval_loss": 0.24446149170398712,
"eval_runtime": 981.0864,
"eval_samples_per_second": 4.8,
"eval_steps_per_second": 0.151,
"eval_wer": 43.091095189355165,
"step": 1000
},
{
"epoch": 6.03,
"learning_rate": 8.83777777777778e-06,
"loss": 0.0047,
"step": 1025
},
{
"epoch": 7.0,
"learning_rate": 8.782222222222223e-06,
"loss": 0.0029,
"step": 1050
},
{
"epoch": 7.01,
"learning_rate": 8.726666666666667e-06,
"loss": 0.0036,
"step": 1075
},
{
"epoch": 7.01,
"learning_rate": 8.671111111111113e-06,
"loss": 0.0027,
"step": 1100
},
{
"epoch": 7.02,
"learning_rate": 8.615555555555555e-06,
"loss": 0.0028,
"step": 1125
},
{
"epoch": 7.02,
"learning_rate": 8.560000000000001e-06,
"loss": 0.0033,
"step": 1150
},
{
"epoch": 7.03,
"learning_rate": 8.504444444444445e-06,
"loss": 0.0039,
"step": 1175
},
{
"epoch": 8.0,
"learning_rate": 8.448888888888889e-06,
"loss": 0.0032,
"step": 1200
},
{
"epoch": 8.01,
"learning_rate": 8.393333333333335e-06,
"loss": 0.0026,
"step": 1225
},
{
"epoch": 8.01,
"learning_rate": 8.337777777777777e-06,
"loss": 0.0018,
"step": 1250
},
{
"epoch": 8.02,
"learning_rate": 8.282222222222223e-06,
"loss": 0.0023,
"step": 1275
},
{
"epoch": 8.02,
"learning_rate": 8.226666666666667e-06,
"loss": 0.0029,
"step": 1300
},
{
"epoch": 8.03,
"learning_rate": 8.171111111111113e-06,
"loss": 0.0024,
"step": 1325
},
{
"epoch": 9.0,
"learning_rate": 8.115555555555557e-06,
"loss": 0.0027,
"step": 1350
},
{
"epoch": 9.01,
"learning_rate": 8.06e-06,
"loss": 0.0013,
"step": 1375
},
{
"epoch": 9.01,
"learning_rate": 8.004444444444445e-06,
"loss": 0.0016,
"step": 1400
},
{
"epoch": 9.02,
"learning_rate": 7.948888888888889e-06,
"loss": 0.0014,
"step": 1425
},
{
"epoch": 9.02,
"learning_rate": 7.893333333333335e-06,
"loss": 0.0023,
"step": 1450
},
{
"epoch": 9.03,
"learning_rate": 7.837777777777779e-06,
"loss": 0.0016,
"step": 1475
},
{
"epoch": 10.0,
"learning_rate": 7.782222222222223e-06,
"loss": 0.0012,
"step": 1500
},
{
"epoch": 10.01,
"learning_rate": 7.726666666666667e-06,
"loss": 0.0013,
"step": 1525
},
{
"epoch": 10.01,
"learning_rate": 7.67111111111111e-06,
"loss": 0.001,
"step": 1550
},
{
"epoch": 10.02,
"learning_rate": 7.6155555555555564e-06,
"loss": 0.001,
"step": 1575
},
{
"epoch": 10.02,
"learning_rate": 7.5600000000000005e-06,
"loss": 0.0015,
"step": 1600
},
{
"epoch": 10.03,
"learning_rate": 7.504444444444445e-06,
"loss": 0.001,
"step": 1625
},
{
"epoch": 11.0,
"learning_rate": 7.44888888888889e-06,
"loss": 0.0009,
"step": 1650
},
{
"epoch": 11.01,
"learning_rate": 7.393333333333333e-06,
"loss": 0.0008,
"step": 1675
},
{
"epoch": 11.01,
"learning_rate": 7.337777777777778e-06,
"loss": 0.001,
"step": 1700
},
{
"epoch": 11.02,
"learning_rate": 7.282222222222222e-06,
"loss": 0.0008,
"step": 1725
},
{
"epoch": 11.02,
"learning_rate": 7.226666666666667e-06,
"loss": 0.001,
"step": 1750
},
{
"epoch": 11.03,
"learning_rate": 7.171111111111112e-06,
"loss": 0.0009,
"step": 1775
},
{
"epoch": 12.0,
"learning_rate": 7.115555555555557e-06,
"loss": 0.0007,
"step": 1800
},
{
"epoch": 12.01,
"learning_rate": 7.06e-06,
"loss": 0.0007,
"step": 1825
},
{
"epoch": 12.01,
"learning_rate": 7.004444444444445e-06,
"loss": 0.0008,
"step": 1850
},
{
"epoch": 12.02,
"learning_rate": 6.948888888888889e-06,
"loss": 0.0007,
"step": 1875
},
{
"epoch": 12.02,
"learning_rate": 6.893333333333334e-06,
"loss": 0.0013,
"step": 1900
},
{
"epoch": 13.0,
"learning_rate": 6.837777777777779e-06,
"loss": 0.0007,
"step": 1925
},
{
"epoch": 13.01,
"learning_rate": 6.782222222222222e-06,
"loss": 0.0006,
"step": 1950
},
{
"epoch": 13.01,
"learning_rate": 6.726666666666667e-06,
"loss": 0.0006,
"step": 1975
},
{
"epoch": 13.02,
"learning_rate": 6.671111111111112e-06,
"loss": 0.0006,
"step": 2000
},
{
"epoch": 13.02,
"eval_loss": 0.2639440596103668,
"eval_runtime": 993.6601,
"eval_samples_per_second": 4.739,
"eval_steps_per_second": 0.149,
"eval_wer": 42.988741044012286,
"step": 2000
},
{
"epoch": 13.02,
"learning_rate": 6.615555555555556e-06,
"loss": 0.0006,
"step": 2025
},
{
"epoch": 13.03,
"learning_rate": 6.560000000000001e-06,
"loss": 0.0006,
"step": 2050
},
{
"epoch": 14.0,
"learning_rate": 6.504444444444446e-06,
"loss": 0.0006,
"step": 2075
},
{
"epoch": 14.01,
"learning_rate": 6.448888888888889e-06,
"loss": 0.0006,
"step": 2100
},
{
"epoch": 14.01,
"learning_rate": 6.393333333333334e-06,
"loss": 0.0005,
"step": 2125
},
{
"epoch": 14.02,
"learning_rate": 6.3377777777777786e-06,
"loss": 0.0005,
"step": 2150
},
{
"epoch": 14.02,
"learning_rate": 6.282222222222223e-06,
"loss": 0.0005,
"step": 2175
},
{
"epoch": 14.03,
"learning_rate": 6.2266666666666675e-06,
"loss": 0.0008,
"step": 2200
},
{
"epoch": 15.0,
"learning_rate": 6.171111111111112e-06,
"loss": 0.0006,
"step": 2225
},
{
"epoch": 15.01,
"learning_rate": 6.1155555555555555e-06,
"loss": 0.0005,
"step": 2250
},
{
"epoch": 15.01,
"learning_rate": 6.0600000000000004e-06,
"loss": 0.0005,
"step": 2275
},
{
"epoch": 15.02,
"learning_rate": 6.004444444444445e-06,
"loss": 0.0005,
"step": 2300
},
{
"epoch": 15.02,
"learning_rate": 5.948888888888889e-06,
"loss": 0.0008,
"step": 2325
},
{
"epoch": 15.03,
"learning_rate": 5.893333333333334e-06,
"loss": 0.0005,
"step": 2350
},
{
"epoch": 16.0,
"learning_rate": 5.837777777777777e-06,
"loss": 0.0005,
"step": 2375
},
{
"epoch": 16.01,
"learning_rate": 5.782222222222222e-06,
"loss": 0.0005,
"step": 2400
},
{
"epoch": 16.01,
"learning_rate": 5.726666666666667e-06,
"loss": 0.0005,
"step": 2425
},
{
"epoch": 16.02,
"learning_rate": 5.671111111111112e-06,
"loss": 0.0005,
"step": 2450
},
{
"epoch": 16.02,
"learning_rate": 5.615555555555556e-06,
"loss": 0.0004,
"step": 2475
},
{
"epoch": 16.03,
"learning_rate": 5.560000000000001e-06,
"loss": 0.0005,
"step": 2500
},
{
"epoch": 17.0,
"learning_rate": 5.504444444444444e-06,
"loss": 0.0004,
"step": 2525
},
{
"epoch": 17.01,
"learning_rate": 5.448888888888889e-06,
"loss": 0.0004,
"step": 2550
},
{
"epoch": 17.01,
"learning_rate": 5.393333333333334e-06,
"loss": 0.0004,
"step": 2575
},
{
"epoch": 17.02,
"learning_rate": 5.337777777777779e-06,
"loss": 0.0004,
"step": 2600
},
{
"epoch": 17.02,
"learning_rate": 5.282222222222223e-06,
"loss": 0.0004,
"step": 2625
},
{
"epoch": 17.03,
"learning_rate": 5.226666666666667e-06,
"loss": 0.0004,
"step": 2650
},
{
"epoch": 18.0,
"learning_rate": 5.171111111111111e-06,
"loss": 0.0004,
"step": 2675
},
{
"epoch": 18.01,
"learning_rate": 5.115555555555556e-06,
"loss": 0.0004,
"step": 2700
},
{
"epoch": 18.01,
"learning_rate": 5.060000000000001e-06,
"loss": 0.0004,
"step": 2725
},
{
"epoch": 18.02,
"learning_rate": 5.004444444444445e-06,
"loss": 0.0004,
"step": 2750
},
{
"epoch": 18.02,
"learning_rate": 4.94888888888889e-06,
"loss": 0.0004,
"step": 2775
},
{
"epoch": 18.03,
"learning_rate": 4.893333333333334e-06,
"loss": 0.0004,
"step": 2800
},
{
"epoch": 19.0,
"learning_rate": 4.837777777777778e-06,
"loss": 0.0004,
"step": 2825
},
{
"epoch": 19.01,
"learning_rate": 4.7822222222222226e-06,
"loss": 0.0004,
"step": 2850
},
{
"epoch": 19.01,
"learning_rate": 4.7266666666666674e-06,
"loss": 0.0003,
"step": 2875
},
{
"epoch": 19.02,
"learning_rate": 4.6711111111111115e-06,
"loss": 0.0003,
"step": 2900
},
{
"epoch": 19.02,
"learning_rate": 4.6155555555555555e-06,
"loss": 0.0003,
"step": 2925
},
{
"epoch": 19.03,
"learning_rate": 4.56e-06,
"loss": 0.0004,
"step": 2950
},
{
"epoch": 20.0,
"learning_rate": 4.504444444444444e-06,
"loss": 0.0003,
"step": 2975
},
{
"epoch": 20.01,
"learning_rate": 4.448888888888889e-06,
"loss": 0.0003,
"step": 3000
},
{
"epoch": 20.01,
"eval_loss": 0.2786634862422943,
"eval_runtime": 973.9213,
"eval_samples_per_second": 4.835,
"eval_steps_per_second": 0.152,
"eval_wer": 43.19344933469805,
"step": 3000
},
{
"epoch": 20.01,
"learning_rate": 4.393333333333334e-06,
"loss": 0.0003,
"step": 3025
},
{
"epoch": 20.02,
"learning_rate": 4.337777777777778e-06,
"loss": 0.0003,
"step": 3050
},
{
"epoch": 20.02,
"learning_rate": 4.282222222222222e-06,
"loss": 0.0003,
"step": 3075
},
{
"epoch": 20.03,
"learning_rate": 4.226666666666667e-06,
"loss": 0.0003,
"step": 3100
},
{
"epoch": 21.0,
"learning_rate": 4.171111111111111e-06,
"loss": 0.0003,
"step": 3125
},
{
"epoch": 21.01,
"learning_rate": 4.115555555555556e-06,
"loss": 0.0003,
"step": 3150
},
{
"epoch": 21.01,
"learning_rate": 4.060000000000001e-06,
"loss": 0.0003,
"step": 3175
},
{
"epoch": 21.02,
"learning_rate": 4.004444444444445e-06,
"loss": 0.0003,
"step": 3200
},
{
"epoch": 21.02,
"learning_rate": 3.948888888888889e-06,
"loss": 0.0003,
"step": 3225
},
{
"epoch": 21.03,
"learning_rate": 3.893333333333333e-06,
"loss": 0.0003,
"step": 3250
},
{
"epoch": 22.0,
"learning_rate": 3.837777777777778e-06,
"loss": 0.0003,
"step": 3275
},
{
"epoch": 22.01,
"learning_rate": 3.782222222222223e-06,
"loss": 0.0003,
"step": 3300
},
{
"epoch": 22.01,
"learning_rate": 3.726666666666667e-06,
"loss": 0.0003,
"step": 3325
},
{
"epoch": 22.02,
"learning_rate": 3.6711111111111113e-06,
"loss": 0.0003,
"step": 3350
},
{
"epoch": 22.02,
"learning_rate": 3.615555555555556e-06,
"loss": 0.0003,
"step": 3375
},
{
"epoch": 22.03,
"learning_rate": 3.5600000000000002e-06,
"loss": 0.0003,
"step": 3400
},
{
"epoch": 23.0,
"learning_rate": 3.5044444444444447e-06,
"loss": 0.0003,
"step": 3425
},
{
"epoch": 23.01,
"learning_rate": 3.4488888888888896e-06,
"loss": 0.0003,
"step": 3450
},
{
"epoch": 23.01,
"learning_rate": 3.3933333333333336e-06,
"loss": 0.0003,
"step": 3475
},
{
"epoch": 23.02,
"learning_rate": 3.337777777777778e-06,
"loss": 0.0003,
"step": 3500
},
{
"epoch": 23.02,
"learning_rate": 3.282222222222223e-06,
"loss": 0.0003,
"step": 3525
},
{
"epoch": 23.03,
"learning_rate": 3.226666666666667e-06,
"loss": 0.0003,
"step": 3550
},
{
"epoch": 24.0,
"learning_rate": 3.1711111111111114e-06,
"loss": 0.0003,
"step": 3575
},
{
"epoch": 24.01,
"learning_rate": 3.1155555555555555e-06,
"loss": 0.0003,
"step": 3600
},
{
"epoch": 24.01,
"learning_rate": 3.0600000000000003e-06,
"loss": 0.0003,
"step": 3625
},
{
"epoch": 24.02,
"learning_rate": 3.004444444444445e-06,
"loss": 0.0003,
"step": 3650
},
{
"epoch": 24.02,
"learning_rate": 2.948888888888889e-06,
"loss": 0.0003,
"step": 3675
},
{
"epoch": 24.03,
"learning_rate": 2.8933333333333337e-06,
"loss": 0.0003,
"step": 3700
},
{
"epoch": 25.0,
"learning_rate": 2.837777777777778e-06,
"loss": 0.0002,
"step": 3725
},
{
"epoch": 25.01,
"learning_rate": 2.7822222222222222e-06,
"loss": 0.0002,
"step": 3750
},
{
"epoch": 25.02,
"learning_rate": 2.726666666666667e-06,
"loss": 0.0002,
"step": 3775
},
{
"epoch": 25.02,
"learning_rate": 2.6711111111111116e-06,
"loss": 0.0002,
"step": 3800
},
{
"epoch": 25.02,
"learning_rate": 2.6155555555555556e-06,
"loss": 0.0002,
"step": 3825
},
{
"epoch": 26.0,
"learning_rate": 2.56e-06,
"loss": 0.0002,
"step": 3850
},
{
"epoch": 26.01,
"learning_rate": 2.504444444444445e-06,
"loss": 0.0002,
"step": 3875
},
{
"epoch": 26.01,
"learning_rate": 2.448888888888889e-06,
"loss": 0.0002,
"step": 3900
},
{
"epoch": 26.02,
"learning_rate": 2.3933333333333334e-06,
"loss": 0.0002,
"step": 3925
},
{
"epoch": 26.02,
"learning_rate": 2.337777777777778e-06,
"loss": 0.0002,
"step": 3950
},
{
"epoch": 26.03,
"learning_rate": 2.2822222222222223e-06,
"loss": 0.0002,
"step": 3975
},
{
"epoch": 27.0,
"learning_rate": 2.226666666666667e-06,
"loss": 0.0002,
"step": 4000
},
{
"epoch": 27.0,
"eval_loss": 0.2877311706542969,
"eval_runtime": 983.8399,
"eval_samples_per_second": 4.786,
"eval_steps_per_second": 0.15,
"eval_wer": 43.54145342886387,
"step": 4000
},
{
"epoch": 27.01,
"learning_rate": 2.1711111111111113e-06,
"loss": 0.0002,
"step": 4025
},
{
"epoch": 27.01,
"learning_rate": 2.1155555555555557e-06,
"loss": 0.0002,
"step": 4050
},
{
"epoch": 27.02,
"learning_rate": 2.06e-06,
"loss": 0.0002,
"step": 4075
},
{
"epoch": 27.02,
"learning_rate": 2.0044444444444446e-06,
"loss": 0.0002,
"step": 4100
},
{
"epoch": 27.03,
"learning_rate": 1.948888888888889e-06,
"loss": 0.0002,
"step": 4125
},
{
"epoch": 28.0,
"learning_rate": 1.8933333333333333e-06,
"loss": 0.0002,
"step": 4150
},
{
"epoch": 28.01,
"learning_rate": 1.837777777777778e-06,
"loss": 0.0002,
"step": 4175
},
{
"epoch": 28.01,
"learning_rate": 1.7822222222222225e-06,
"loss": 0.0002,
"step": 4200
},
{
"epoch": 28.02,
"learning_rate": 1.7266666666666667e-06,
"loss": 0.0002,
"step": 4225
},
{
"epoch": 28.02,
"learning_rate": 1.6711111111111112e-06,
"loss": 0.0002,
"step": 4250
},
{
"epoch": 28.03,
"learning_rate": 1.6155555555555559e-06,
"loss": 0.0002,
"step": 4275
},
{
"epoch": 29.0,
"learning_rate": 1.56e-06,
"loss": 0.0002,
"step": 4300
},
{
"epoch": 29.01,
"learning_rate": 1.5044444444444446e-06,
"loss": 0.0002,
"step": 4325
},
{
"epoch": 29.01,
"learning_rate": 1.4488888888888892e-06,
"loss": 0.0002,
"step": 4350
},
{
"epoch": 29.02,
"learning_rate": 1.3933333333333335e-06,
"loss": 0.0002,
"step": 4375
},
{
"epoch": 29.02,
"learning_rate": 1.337777777777778e-06,
"loss": 0.0002,
"step": 4400
},
{
"epoch": 29.03,
"learning_rate": 1.2822222222222222e-06,
"loss": 0.0002,
"step": 4425
},
{
"epoch": 30.0,
"learning_rate": 1.2266666666666666e-06,
"loss": 0.0002,
"step": 4450
},
{
"epoch": 30.01,
"learning_rate": 1.171111111111111e-06,
"loss": 0.0002,
"step": 4475
},
{
"epoch": 30.01,
"learning_rate": 1.1155555555555558e-06,
"loss": 0.0002,
"step": 4500
},
{
"epoch": 30.02,
"learning_rate": 1.06e-06,
"loss": 0.0002,
"step": 4525
},
{
"epoch": 30.02,
"learning_rate": 1.0044444444444445e-06,
"loss": 0.0002,
"step": 4550
},
{
"epoch": 30.03,
"learning_rate": 9.488888888888889e-07,
"loss": 0.0002,
"step": 4575
},
{
"epoch": 31.0,
"learning_rate": 8.933333333333334e-07,
"loss": 0.0002,
"step": 4600
},
{
"epoch": 31.01,
"learning_rate": 8.37777777777778e-07,
"loss": 0.0002,
"step": 4625
},
{
"epoch": 31.01,
"learning_rate": 7.822222222222223e-07,
"loss": 0.0002,
"step": 4650
},
{
"epoch": 31.02,
"learning_rate": 7.266666666666668e-07,
"loss": 0.0002,
"step": 4675
},
{
"epoch": 31.02,
"learning_rate": 6.711111111111111e-07,
"loss": 0.0002,
"step": 4700
},
{
"epoch": 31.03,
"learning_rate": 6.155555555555556e-07,
"loss": 0.0002,
"step": 4725
},
{
"epoch": 32.0,
"learning_rate": 5.6e-07,
"loss": 0.0002,
"step": 4750
},
{
"epoch": 32.01,
"learning_rate": 5.044444444444445e-07,
"loss": 0.0002,
"step": 4775
},
{
"epoch": 32.01,
"learning_rate": 4.488888888888889e-07,
"loss": 0.0002,
"step": 4800
},
{
"epoch": 32.02,
"learning_rate": 3.9333333333333336e-07,
"loss": 0.0002,
"step": 4825
},
{
"epoch": 32.02,
"learning_rate": 3.3777777777777777e-07,
"loss": 0.0002,
"step": 4850
},
{
"epoch": 32.03,
"learning_rate": 2.822222222222222e-07,
"loss": 0.0002,
"step": 4875
},
{
"epoch": 33.0,
"learning_rate": 2.266666666666667e-07,
"loss": 0.0002,
"step": 4900
},
{
"epoch": 33.01,
"learning_rate": 1.7111111111111114e-07,
"loss": 0.0002,
"step": 4925
},
{
"epoch": 33.01,
"learning_rate": 1.1555555555555556e-07,
"loss": 0.0002,
"step": 4950
},
{
"epoch": 33.02,
"learning_rate": 6.000000000000001e-08,
"loss": 0.0002,
"step": 4975
},
{
"epoch": 33.02,
"learning_rate": 4.444444444444445e-09,
"loss": 0.0002,
"step": 5000
},
{
"epoch": 33.02,
"eval_loss": 0.29104799032211304,
"eval_runtime": 967.1931,
"eval_samples_per_second": 4.869,
"eval_steps_per_second": 0.153,
"eval_wer": 43.582395087001025,
"step": 5000
},
{
"epoch": 33.02,
"step": 5000,
"total_flos": 9.183306891902976e+19,
"train_loss": 0.027913081424683333,
"train_runtime": 38085.1906,
"train_samples_per_second": 8.402,
"train_steps_per_second": 0.131
}
],
"max_steps": 5000,
"num_train_epochs": 9223372036854775807,
"total_flos": 9.183306891902976e+19,
"trial_name": null,
"trial_params": null
}