whisper-tiny-it / trainer_state.json
mattiasu96's picture
End of training
2097873
raw
history blame
No virus
25.7 kB
{
"best_metric": 26.495056347012547,
"best_model_checkpoint": "./checkpoint-5000",
"epoch": 1.046,
"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": 4.4e-07,
"loss": 1.5447,
"step": 25
},
{
"epoch": 0.01,
"learning_rate": 9.400000000000001e-07,
"loss": 1.3096,
"step": 50
},
{
"epoch": 0.01,
"learning_rate": 1.44e-06,
"loss": 1.2002,
"step": 75
},
{
"epoch": 0.02,
"learning_rate": 1.94e-06,
"loss": 1.0145,
"step": 100
},
{
"epoch": 0.03,
"learning_rate": 2.4400000000000004e-06,
"loss": 0.837,
"step": 125
},
{
"epoch": 0.03,
"learning_rate": 2.9400000000000002e-06,
"loss": 0.7799,
"step": 150
},
{
"epoch": 0.04,
"learning_rate": 3.44e-06,
"loss": 0.7134,
"step": 175
},
{
"epoch": 0.04,
"learning_rate": 3.94e-06,
"loss": 0.9371,
"step": 200
},
{
"epoch": 0.04,
"learning_rate": 4.440000000000001e-06,
"loss": 0.8885,
"step": 225
},
{
"epoch": 0.05,
"learning_rate": 4.94e-06,
"loss": 0.8197,
"step": 250
},
{
"epoch": 0.06,
"learning_rate": 5.4400000000000004e-06,
"loss": 0.7636,
"step": 275
},
{
"epoch": 0.06,
"learning_rate": 5.94e-06,
"loss": 0.7776,
"step": 300
},
{
"epoch": 0.07,
"learning_rate": 6.440000000000001e-06,
"loss": 0.9414,
"step": 325
},
{
"epoch": 0.07,
"learning_rate": 6.9400000000000005e-06,
"loss": 0.8638,
"step": 350
},
{
"epoch": 0.07,
"learning_rate": 7.440000000000001e-06,
"loss": 0.8138,
"step": 375
},
{
"epoch": 0.08,
"learning_rate": 7.94e-06,
"loss": 0.8551,
"step": 400
},
{
"epoch": 0.09,
"learning_rate": 8.44e-06,
"loss": 0.753,
"step": 425
},
{
"epoch": 0.09,
"learning_rate": 8.94e-06,
"loss": 0.5542,
"step": 450
},
{
"epoch": 0.1,
"learning_rate": 9.440000000000001e-06,
"loss": 0.5019,
"step": 475
},
{
"epoch": 0.1,
"learning_rate": 9.940000000000001e-06,
"loss": 0.5859,
"step": 500
},
{
"epoch": 0.1,
"learning_rate": 9.951111111111111e-06,
"loss": 0.5963,
"step": 525
},
{
"epoch": 0.11,
"learning_rate": 9.895555555555557e-06,
"loss": 0.558,
"step": 550
},
{
"epoch": 0.12,
"learning_rate": 9.84e-06,
"loss": 0.5404,
"step": 575
},
{
"epoch": 0.12,
"learning_rate": 9.784444444444445e-06,
"loss": 0.5436,
"step": 600
},
{
"epoch": 0.12,
"learning_rate": 9.72888888888889e-06,
"loss": 0.5207,
"step": 625
},
{
"epoch": 0.13,
"learning_rate": 9.673333333333334e-06,
"loss": 0.5452,
"step": 650
},
{
"epoch": 0.14,
"learning_rate": 9.617777777777778e-06,
"loss": 0.5679,
"step": 675
},
{
"epoch": 0.14,
"learning_rate": 9.562222222222223e-06,
"loss": 0.5116,
"step": 700
},
{
"epoch": 0.14,
"learning_rate": 9.506666666666667e-06,
"loss": 0.5101,
"step": 725
},
{
"epoch": 0.15,
"learning_rate": 9.451111111111112e-06,
"loss": 0.46,
"step": 750
},
{
"epoch": 0.15,
"learning_rate": 9.395555555555556e-06,
"loss": 0.4566,
"step": 775
},
{
"epoch": 0.16,
"learning_rate": 9.340000000000002e-06,
"loss": 0.4505,
"step": 800
},
{
"epoch": 0.17,
"learning_rate": 9.284444444444444e-06,
"loss": 0.4622,
"step": 825
},
{
"epoch": 0.17,
"learning_rate": 9.22888888888889e-06,
"loss": 0.4355,
"step": 850
},
{
"epoch": 0.17,
"learning_rate": 9.173333333333334e-06,
"loss": 0.4457,
"step": 875
},
{
"epoch": 0.18,
"learning_rate": 9.117777777777778e-06,
"loss": 0.4864,
"step": 900
},
{
"epoch": 0.18,
"learning_rate": 9.062222222222224e-06,
"loss": 0.4873,
"step": 925
},
{
"epoch": 0.19,
"learning_rate": 9.006666666666666e-06,
"loss": 0.4953,
"step": 950
},
{
"epoch": 0.2,
"learning_rate": 8.951111111111112e-06,
"loss": 0.4615,
"step": 975
},
{
"epoch": 0.2,
"learning_rate": 8.895555555555556e-06,
"loss": 0.3889,
"step": 1000
},
{
"epoch": 0.2,
"eval_loss": 0.6844463348388672,
"eval_runtime": 3611.4391,
"eval_samples_per_second": 4.154,
"eval_steps_per_second": 0.13,
"eval_wer": 39.100042526047204,
"step": 1000
},
{
"epoch": 0.2,
"learning_rate": 8.84e-06,
"loss": 0.4443,
"step": 1025
},
{
"epoch": 0.21,
"learning_rate": 8.784444444444446e-06,
"loss": 0.4519,
"step": 1050
},
{
"epoch": 0.21,
"learning_rate": 8.72888888888889e-06,
"loss": 0.4655,
"step": 1075
},
{
"epoch": 0.22,
"learning_rate": 8.673333333333334e-06,
"loss": 0.47,
"step": 1100
},
{
"epoch": 0.23,
"learning_rate": 8.617777777777778e-06,
"loss": 0.4648,
"step": 1125
},
{
"epoch": 0.23,
"learning_rate": 8.562222222222224e-06,
"loss": 0.464,
"step": 1150
},
{
"epoch": 0.23,
"learning_rate": 8.506666666666668e-06,
"loss": 0.5185,
"step": 1175
},
{
"epoch": 0.24,
"learning_rate": 8.451111111111112e-06,
"loss": 0.5426,
"step": 1200
},
{
"epoch": 0.24,
"learning_rate": 8.395555555555557e-06,
"loss": 0.4005,
"step": 1225
},
{
"epoch": 0.25,
"learning_rate": 8.34e-06,
"loss": 0.4603,
"step": 1250
},
{
"epoch": 0.26,
"learning_rate": 8.284444444444446e-06,
"loss": 0.4283,
"step": 1275
},
{
"epoch": 0.26,
"learning_rate": 8.22888888888889e-06,
"loss": 0.4383,
"step": 1300
},
{
"epoch": 0.27,
"learning_rate": 8.173333333333334e-06,
"loss": 0.4516,
"step": 1325
},
{
"epoch": 0.27,
"learning_rate": 8.11777777777778e-06,
"loss": 0.4893,
"step": 1350
},
{
"epoch": 0.28,
"learning_rate": 8.062222222222222e-06,
"loss": 0.517,
"step": 1375
},
{
"epoch": 0.28,
"learning_rate": 8.006666666666667e-06,
"loss": 0.4487,
"step": 1400
},
{
"epoch": 0.28,
"learning_rate": 7.951111111111111e-06,
"loss": 0.4357,
"step": 1425
},
{
"epoch": 0.29,
"learning_rate": 7.895555555555557e-06,
"loss": 0.417,
"step": 1450
},
{
"epoch": 0.29,
"learning_rate": 7.840000000000001e-06,
"loss": 0.4386,
"step": 1475
},
{
"epoch": 0.3,
"learning_rate": 7.784444444444445e-06,
"loss": 0.3935,
"step": 1500
},
{
"epoch": 0.3,
"learning_rate": 7.72888888888889e-06,
"loss": 0.3887,
"step": 1525
},
{
"epoch": 0.31,
"learning_rate": 7.673333333333333e-06,
"loss": 0.4385,
"step": 1550
},
{
"epoch": 0.32,
"learning_rate": 7.617777777777778e-06,
"loss": 0.5525,
"step": 1575
},
{
"epoch": 0.32,
"learning_rate": 7.562222222222223e-06,
"loss": 0.4988,
"step": 1600
},
{
"epoch": 0.33,
"learning_rate": 7.506666666666668e-06,
"loss": 0.4345,
"step": 1625
},
{
"epoch": 0.33,
"learning_rate": 7.451111111111111e-06,
"loss": 0.3784,
"step": 1650
},
{
"epoch": 0.34,
"learning_rate": 7.395555555555556e-06,
"loss": 0.3756,
"step": 1675
},
{
"epoch": 0.34,
"learning_rate": 7.340000000000001e-06,
"loss": 0.3977,
"step": 1700
},
{
"epoch": 0.34,
"learning_rate": 7.284444444444445e-06,
"loss": 0.3918,
"step": 1725
},
{
"epoch": 0.35,
"learning_rate": 7.22888888888889e-06,
"loss": 0.4001,
"step": 1750
},
{
"epoch": 0.35,
"learning_rate": 7.173333333333335e-06,
"loss": 0.416,
"step": 1775
},
{
"epoch": 0.36,
"learning_rate": 7.117777777777778e-06,
"loss": 0.4195,
"step": 1800
},
{
"epoch": 0.36,
"learning_rate": 7.062222222222223e-06,
"loss": 0.4248,
"step": 1825
},
{
"epoch": 0.37,
"learning_rate": 7.006666666666667e-06,
"loss": 0.3734,
"step": 1850
},
{
"epoch": 0.38,
"learning_rate": 6.951111111111112e-06,
"loss": 0.4352,
"step": 1875
},
{
"epoch": 0.38,
"learning_rate": 6.8955555555555565e-06,
"loss": 0.6166,
"step": 1900
},
{
"epoch": 0.39,
"learning_rate": 6.8400000000000014e-06,
"loss": 0.7169,
"step": 1925
},
{
"epoch": 0.39,
"learning_rate": 6.784444444444445e-06,
"loss": 0.6759,
"step": 1950
},
{
"epoch": 0.4,
"learning_rate": 6.7288888888888895e-06,
"loss": 0.666,
"step": 1975
},
{
"epoch": 0.4,
"learning_rate": 6.6733333333333335e-06,
"loss": 0.5832,
"step": 2000
},
{
"epoch": 0.4,
"eval_loss": 0.5691020488739014,
"eval_runtime": 3704.1035,
"eval_samples_per_second": 4.05,
"eval_steps_per_second": 0.127,
"eval_wer": 31.999521581968953,
"step": 2000
},
{
"epoch": 0.41,
"learning_rate": 6.617777777777778e-06,
"loss": 0.4642,
"step": 2025
},
{
"epoch": 0.41,
"learning_rate": 6.562222222222223e-06,
"loss": 0.4274,
"step": 2050
},
{
"epoch": 0.41,
"learning_rate": 6.5066666666666665e-06,
"loss": 0.3927,
"step": 2075
},
{
"epoch": 0.42,
"learning_rate": 6.451111111111111e-06,
"loss": 0.3872,
"step": 2100
},
{
"epoch": 0.42,
"learning_rate": 6.395555555555556e-06,
"loss": 0.3513,
"step": 2125
},
{
"epoch": 0.43,
"learning_rate": 6.34e-06,
"loss": 0.4006,
"step": 2150
},
{
"epoch": 0.43,
"learning_rate": 6.284444444444445e-06,
"loss": 0.439,
"step": 2175
},
{
"epoch": 0.44,
"learning_rate": 6.22888888888889e-06,
"loss": 0.454,
"step": 2200
},
{
"epoch": 0.45,
"learning_rate": 6.173333333333333e-06,
"loss": 0.4747,
"step": 2225
},
{
"epoch": 0.45,
"learning_rate": 6.117777777777778e-06,
"loss": 0.465,
"step": 2250
},
{
"epoch": 0.46,
"learning_rate": 6.062222222222223e-06,
"loss": 0.5166,
"step": 2275
},
{
"epoch": 0.46,
"learning_rate": 6.006666666666667e-06,
"loss": 0.4957,
"step": 2300
},
{
"epoch": 0.47,
"learning_rate": 5.951111111111112e-06,
"loss": 0.5049,
"step": 2325
},
{
"epoch": 0.47,
"learning_rate": 5.895555555555557e-06,
"loss": 0.532,
"step": 2350
},
{
"epoch": 0.47,
"learning_rate": 5.84e-06,
"loss": 0.4619,
"step": 2375
},
{
"epoch": 0.48,
"learning_rate": 5.784444444444445e-06,
"loss": 0.4376,
"step": 2400
},
{
"epoch": 0.48,
"learning_rate": 5.72888888888889e-06,
"loss": 0.5131,
"step": 2425
},
{
"epoch": 0.49,
"learning_rate": 5.673333333333334e-06,
"loss": 0.4956,
"step": 2450
},
{
"epoch": 0.49,
"learning_rate": 5.617777777777779e-06,
"loss": 0.4794,
"step": 2475
},
{
"epoch": 0.5,
"learning_rate": 5.562222222222222e-06,
"loss": 0.4381,
"step": 2500
},
{
"epoch": 0.51,
"learning_rate": 5.506666666666667e-06,
"loss": 0.4671,
"step": 2525
},
{
"epoch": 0.51,
"learning_rate": 5.451111111111112e-06,
"loss": 0.4625,
"step": 2550
},
{
"epoch": 0.52,
"learning_rate": 5.3955555555555565e-06,
"loss": 0.469,
"step": 2575
},
{
"epoch": 0.52,
"learning_rate": 5.3400000000000005e-06,
"loss": 0.4353,
"step": 2600
},
{
"epoch": 0.53,
"learning_rate": 5.2844444444444454e-06,
"loss": 0.4542,
"step": 2625
},
{
"epoch": 0.53,
"learning_rate": 5.228888888888889e-06,
"loss": 0.4441,
"step": 2650
},
{
"epoch": 0.54,
"learning_rate": 5.1733333333333335e-06,
"loss": 0.4431,
"step": 2675
},
{
"epoch": 0.54,
"learning_rate": 5.117777777777778e-06,
"loss": 0.5278,
"step": 2700
},
{
"epoch": 0.55,
"learning_rate": 5.062222222222222e-06,
"loss": 0.4286,
"step": 2725
},
{
"epoch": 0.55,
"learning_rate": 5.006666666666667e-06,
"loss": 0.4339,
"step": 2750
},
{
"epoch": 0.56,
"learning_rate": 4.951111111111111e-06,
"loss": 0.4345,
"step": 2775
},
{
"epoch": 0.56,
"learning_rate": 4.895555555555556e-06,
"loss": 0.4649,
"step": 2800
},
{
"epoch": 0.56,
"learning_rate": 4.84e-06,
"loss": 0.4715,
"step": 2825
},
{
"epoch": 0.57,
"learning_rate": 4.784444444444445e-06,
"loss": 0.4395,
"step": 2850
},
{
"epoch": 0.57,
"learning_rate": 4.728888888888889e-06,
"loss": 0.4528,
"step": 2875
},
{
"epoch": 0.58,
"learning_rate": 4.673333333333333e-06,
"loss": 0.5037,
"step": 2900
},
{
"epoch": 0.58,
"learning_rate": 4.617777777777778e-06,
"loss": 0.4348,
"step": 2925
},
{
"epoch": 0.59,
"learning_rate": 4.562222222222222e-06,
"loss": 0.3983,
"step": 2950
},
{
"epoch": 0.59,
"learning_rate": 4.506666666666667e-06,
"loss": 0.4518,
"step": 2975
},
{
"epoch": 0.6,
"learning_rate": 4.451111111111112e-06,
"loss": 0.524,
"step": 3000
},
{
"epoch": 0.6,
"eval_loss": 0.4992637634277344,
"eval_runtime": 3744.3181,
"eval_samples_per_second": 4.007,
"eval_steps_per_second": 0.125,
"eval_wer": 28.267196470338078,
"step": 3000
},
{
"epoch": 0.6,
"learning_rate": 4.395555555555556e-06,
"loss": 0.4691,
"step": 3025
},
{
"epoch": 0.61,
"learning_rate": 4.34e-06,
"loss": 0.4773,
"step": 3050
},
{
"epoch": 0.61,
"learning_rate": 4.284444444444445e-06,
"loss": 0.4292,
"step": 3075
},
{
"epoch": 0.62,
"learning_rate": 4.228888888888889e-06,
"loss": 0.4509,
"step": 3100
},
{
"epoch": 0.62,
"learning_rate": 4.173333333333334e-06,
"loss": 0.4616,
"step": 3125
},
{
"epoch": 0.63,
"learning_rate": 4.117777777777779e-06,
"loss": 0.4611,
"step": 3150
},
{
"epoch": 0.64,
"learning_rate": 4.062222222222223e-06,
"loss": 0.4197,
"step": 3175
},
{
"epoch": 0.64,
"learning_rate": 4.006666666666667e-06,
"loss": 0.4221,
"step": 3200
},
{
"epoch": 0.65,
"learning_rate": 3.951111111111112e-06,
"loss": 0.4037,
"step": 3225
},
{
"epoch": 0.65,
"learning_rate": 3.895555555555556e-06,
"loss": 0.4381,
"step": 3250
},
{
"epoch": 0.66,
"learning_rate": 3.8400000000000005e-06,
"loss": 0.4315,
"step": 3275
},
{
"epoch": 0.66,
"learning_rate": 3.784444444444445e-06,
"loss": 0.4196,
"step": 3300
},
{
"epoch": 0.67,
"learning_rate": 3.728888888888889e-06,
"loss": 0.4024,
"step": 3325
},
{
"epoch": 0.67,
"learning_rate": 3.673333333333334e-06,
"loss": 0.4098,
"step": 3350
},
{
"epoch": 0.68,
"learning_rate": 3.617777777777778e-06,
"loss": 0.4326,
"step": 3375
},
{
"epoch": 0.68,
"learning_rate": 3.5622222222222224e-06,
"loss": 0.4228,
"step": 3400
},
{
"epoch": 0.69,
"learning_rate": 3.5066666666666673e-06,
"loss": 0.3965,
"step": 3425
},
{
"epoch": 0.69,
"learning_rate": 3.4511111111111113e-06,
"loss": 0.3812,
"step": 3450
},
{
"epoch": 0.69,
"learning_rate": 3.3955555555555558e-06,
"loss": 0.4192,
"step": 3475
},
{
"epoch": 0.7,
"learning_rate": 3.3400000000000006e-06,
"loss": 0.4566,
"step": 3500
},
{
"epoch": 0.7,
"learning_rate": 3.2844444444444447e-06,
"loss": 0.5236,
"step": 3525
},
{
"epoch": 0.71,
"learning_rate": 3.228888888888889e-06,
"loss": 0.4424,
"step": 3550
},
{
"epoch": 0.71,
"learning_rate": 3.173333333333334e-06,
"loss": 0.5007,
"step": 3575
},
{
"epoch": 0.72,
"learning_rate": 3.117777777777778e-06,
"loss": 0.5105,
"step": 3600
},
{
"epoch": 0.72,
"learning_rate": 3.0622222222222225e-06,
"loss": 0.5139,
"step": 3625
},
{
"epoch": 0.73,
"learning_rate": 3.0066666666666674e-06,
"loss": 0.4169,
"step": 3650
},
{
"epoch": 0.73,
"learning_rate": 2.9511111111111114e-06,
"loss": 0.4356,
"step": 3675
},
{
"epoch": 0.74,
"learning_rate": 2.895555555555556e-06,
"loss": 0.4315,
"step": 3700
},
{
"epoch": 0.74,
"learning_rate": 2.84e-06,
"loss": 0.428,
"step": 3725
},
{
"epoch": 0.75,
"learning_rate": 2.784444444444445e-06,
"loss": 0.4267,
"step": 3750
},
{
"epoch": 0.76,
"learning_rate": 2.7288888888888893e-06,
"loss": 0.4324,
"step": 3775
},
{
"epoch": 0.76,
"learning_rate": 2.6733333333333333e-06,
"loss": 0.4901,
"step": 3800
},
{
"epoch": 0.77,
"learning_rate": 2.617777777777778e-06,
"loss": 0.4488,
"step": 3825
},
{
"epoch": 0.77,
"learning_rate": 2.5622222222222226e-06,
"loss": 0.4215,
"step": 3850
},
{
"epoch": 0.78,
"learning_rate": 2.5066666666666667e-06,
"loss": 0.3727,
"step": 3875
},
{
"epoch": 0.78,
"learning_rate": 2.451111111111111e-06,
"loss": 0.3392,
"step": 3900
},
{
"epoch": 0.79,
"learning_rate": 2.3955555555555556e-06,
"loss": 0.3234,
"step": 3925
},
{
"epoch": 0.79,
"learning_rate": 2.3400000000000005e-06,
"loss": 0.335,
"step": 3950
},
{
"epoch": 0.8,
"learning_rate": 2.2844444444444445e-06,
"loss": 0.4014,
"step": 3975
},
{
"epoch": 0.8,
"learning_rate": 2.228888888888889e-06,
"loss": 0.4663,
"step": 4000
},
{
"epoch": 0.8,
"eval_loss": 0.4799182713031769,
"eval_runtime": 3633.393,
"eval_samples_per_second": 4.129,
"eval_steps_per_second": 0.129,
"eval_wer": 26.722969381246013,
"step": 4000
},
{
"epoch": 0.81,
"learning_rate": 2.1733333333333334e-06,
"loss": 0.4464,
"step": 4025
},
{
"epoch": 0.81,
"learning_rate": 2.12e-06,
"loss": 0.4185,
"step": 4050
},
{
"epoch": 0.81,
"learning_rate": 2.064444444444445e-06,
"loss": 0.4123,
"step": 4075
},
{
"epoch": 0.82,
"learning_rate": 2.008888888888889e-06,
"loss": 0.4612,
"step": 4100
},
{
"epoch": 0.82,
"learning_rate": 1.9533333333333334e-06,
"loss": 0.3796,
"step": 4125
},
{
"epoch": 0.83,
"learning_rate": 1.8977777777777779e-06,
"loss": 0.4257,
"step": 4150
},
{
"epoch": 0.83,
"learning_rate": 1.8422222222222225e-06,
"loss": 0.4634,
"step": 4175
},
{
"epoch": 0.84,
"learning_rate": 1.7866666666666668e-06,
"loss": 0.4579,
"step": 4200
},
{
"epoch": 0.84,
"learning_rate": 1.7311111111111112e-06,
"loss": 0.3981,
"step": 4225
},
{
"epoch": 0.85,
"learning_rate": 1.675555555555556e-06,
"loss": 0.3406,
"step": 4250
},
{
"epoch": 0.85,
"learning_rate": 1.6200000000000002e-06,
"loss": 0.3297,
"step": 4275
},
{
"epoch": 0.86,
"learning_rate": 1.5644444444444446e-06,
"loss": 0.3747,
"step": 4300
},
{
"epoch": 0.86,
"learning_rate": 1.5088888888888889e-06,
"loss": 0.402,
"step": 4325
},
{
"epoch": 0.87,
"learning_rate": 1.4533333333333335e-06,
"loss": 0.4979,
"step": 4350
},
{
"epoch": 0.88,
"learning_rate": 1.397777777777778e-06,
"loss": 0.4554,
"step": 4375
},
{
"epoch": 0.88,
"learning_rate": 1.3422222222222222e-06,
"loss": 0.357,
"step": 4400
},
{
"epoch": 0.89,
"learning_rate": 1.286666666666667e-06,
"loss": 0.3165,
"step": 4425
},
{
"epoch": 0.89,
"learning_rate": 1.2311111111111112e-06,
"loss": 0.3502,
"step": 4450
},
{
"epoch": 0.9,
"learning_rate": 1.1755555555555556e-06,
"loss": 0.5364,
"step": 4475
},
{
"epoch": 0.9,
"learning_rate": 1.12e-06,
"loss": 0.4442,
"step": 4500
},
{
"epoch": 0.91,
"learning_rate": 1.0644444444444445e-06,
"loss": 0.3887,
"step": 4525
},
{
"epoch": 0.91,
"learning_rate": 1.008888888888889e-06,
"loss": 0.3367,
"step": 4550
},
{
"epoch": 0.92,
"learning_rate": 9.533333333333335e-07,
"loss": 0.5158,
"step": 4575
},
{
"epoch": 0.92,
"learning_rate": 8.977777777777778e-07,
"loss": 0.4902,
"step": 4600
},
{
"epoch": 0.93,
"learning_rate": 8.422222222222224e-07,
"loss": 0.4327,
"step": 4625
},
{
"epoch": 0.93,
"learning_rate": 7.888888888888889e-07,
"loss": 0.4298,
"step": 4650
},
{
"epoch": 0.94,
"learning_rate": 7.333333333333334e-07,
"loss": 0.3963,
"step": 4675
},
{
"epoch": 0.94,
"learning_rate": 6.777777777777779e-07,
"loss": 0.3787,
"step": 4700
},
{
"epoch": 0.94,
"learning_rate": 6.222222222222223e-07,
"loss": 0.3294,
"step": 4725
},
{
"epoch": 0.95,
"learning_rate": 5.666666666666667e-07,
"loss": 0.3712,
"step": 4750
},
{
"epoch": 1.0,
"learning_rate": 5.111111111111112e-07,
"loss": 0.3613,
"step": 4775
},
{
"epoch": 1.01,
"learning_rate": 4.5555555555555563e-07,
"loss": 0.2949,
"step": 4800
},
{
"epoch": 1.01,
"learning_rate": 4.0000000000000003e-07,
"loss": 0.2963,
"step": 4825
},
{
"epoch": 1.02,
"learning_rate": 3.444444444444445e-07,
"loss": 0.3185,
"step": 4850
},
{
"epoch": 1.02,
"learning_rate": 2.888888888888889e-07,
"loss": 0.2978,
"step": 4875
},
{
"epoch": 1.03,
"learning_rate": 2.3333333333333336e-07,
"loss": 0.3146,
"step": 4900
},
{
"epoch": 1.03,
"learning_rate": 1.777777777777778e-07,
"loss": 0.3337,
"step": 4925
},
{
"epoch": 1.04,
"learning_rate": 1.2222222222222225e-07,
"loss": 0.3924,
"step": 4950
},
{
"epoch": 1.04,
"learning_rate": 6.666666666666668e-08,
"loss": 0.2629,
"step": 4975
},
{
"epoch": 1.05,
"learning_rate": 1.1111111111111112e-08,
"loss": 0.3107,
"step": 5000
},
{
"epoch": 1.05,
"eval_loss": 0.47362709045410156,
"eval_runtime": 3770.6714,
"eval_samples_per_second": 3.979,
"eval_steps_per_second": 0.124,
"eval_wer": 26.495056347012547,
"step": 5000
},
{
"epoch": 1.05,
"step": 5000,
"total_flos": 3.93825843376128e+18,
"train_loss": 0.485806952381134,
"train_runtime": 42089.9231,
"train_samples_per_second": 3.801,
"train_steps_per_second": 0.119
}
],
"max_steps": 5000,
"num_train_epochs": 9223372036854775807,
"total_flos": 3.93825843376128e+18,
"trial_name": null,
"trial_params": null
}