whisper-el-medium-augmented / trainer_state.json
alkiskoudounas's picture
End of training
e6d089d
{
"best_metric": 17.09695393759287,
"best_model_checkpoint": "whisper-el-medium-augmented-2/checkpoint-20000",
"epoch": 46.948356807511736,
"global_step": 20000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.12,
"learning_rate": 9.400000000000001e-07,
"loss": 2.3145,
"step": 50
},
{
"epoch": 0.23,
"learning_rate": 1.94e-06,
"loss": 1.4804,
"step": 100
},
{
"epoch": 0.35,
"learning_rate": 2.9400000000000002e-06,
"loss": 0.6645,
"step": 150
},
{
"epoch": 0.47,
"learning_rate": 3.94e-06,
"loss": 0.4941,
"step": 200
},
{
"epoch": 0.59,
"learning_rate": 4.94e-06,
"loss": 0.4274,
"step": 250
},
{
"epoch": 0.7,
"learning_rate": 5.94e-06,
"loss": 0.3706,
"step": 300
},
{
"epoch": 0.82,
"learning_rate": 6.9400000000000005e-06,
"loss": 0.3363,
"step": 350
},
{
"epoch": 0.94,
"learning_rate": 7.94e-06,
"loss": 0.3374,
"step": 400
},
{
"epoch": 1.06,
"learning_rate": 8.94e-06,
"loss": 0.25,
"step": 450
},
{
"epoch": 1.17,
"learning_rate": 9.940000000000001e-06,
"loss": 0.2338,
"step": 500
},
{
"epoch": 1.29,
"learning_rate": 9.975897435897436e-06,
"loss": 0.2485,
"step": 550
},
{
"epoch": 1.41,
"learning_rate": 9.950256410256412e-06,
"loss": 0.2306,
"step": 600
},
{
"epoch": 1.53,
"learning_rate": 9.924615384615385e-06,
"loss": 0.2184,
"step": 650
},
{
"epoch": 1.64,
"learning_rate": 9.89897435897436e-06,
"loss": 0.2268,
"step": 700
},
{
"epoch": 1.76,
"learning_rate": 9.873333333333334e-06,
"loss": 0.1912,
"step": 750
},
{
"epoch": 1.88,
"learning_rate": 9.847692307692308e-06,
"loss": 0.1835,
"step": 800
},
{
"epoch": 2.0,
"learning_rate": 9.822051282051283e-06,
"loss": 0.1908,
"step": 850
},
{
"epoch": 2.11,
"learning_rate": 9.796410256410257e-06,
"loss": 0.1107,
"step": 900
},
{
"epoch": 2.23,
"learning_rate": 9.770769230769232e-06,
"loss": 0.1117,
"step": 950
},
{
"epoch": 2.35,
"learning_rate": 9.745128205128206e-06,
"loss": 0.0995,
"step": 1000
},
{
"epoch": 2.46,
"learning_rate": 9.71948717948718e-06,
"loss": 0.1149,
"step": 1050
},
{
"epoch": 2.58,
"learning_rate": 9.693846153846155e-06,
"loss": 0.1002,
"step": 1100
},
{
"epoch": 2.7,
"learning_rate": 9.668205128205129e-06,
"loss": 0.1142,
"step": 1150
},
{
"epoch": 2.82,
"learning_rate": 9.642564102564104e-06,
"loss": 0.1061,
"step": 1200
},
{
"epoch": 2.93,
"learning_rate": 9.616923076923077e-06,
"loss": 0.1017,
"step": 1250
},
{
"epoch": 3.05,
"learning_rate": 9.591282051282053e-06,
"loss": 0.0797,
"step": 1300
},
{
"epoch": 3.17,
"learning_rate": 9.565641025641026e-06,
"loss": 0.0545,
"step": 1350
},
{
"epoch": 3.29,
"learning_rate": 9.54e-06,
"loss": 0.0548,
"step": 1400
},
{
"epoch": 3.4,
"learning_rate": 9.514358974358975e-06,
"loss": 0.0638,
"step": 1450
},
{
"epoch": 3.52,
"learning_rate": 9.488717948717949e-06,
"loss": 0.0658,
"step": 1500
},
{
"epoch": 3.64,
"learning_rate": 9.463076923076924e-06,
"loss": 0.0674,
"step": 1550
},
{
"epoch": 3.76,
"learning_rate": 9.437435897435898e-06,
"loss": 0.0534,
"step": 1600
},
{
"epoch": 3.87,
"learning_rate": 9.411794871794872e-06,
"loss": 0.0806,
"step": 1650
},
{
"epoch": 3.99,
"learning_rate": 9.386153846153847e-06,
"loss": 0.0559,
"step": 1700
},
{
"epoch": 4.11,
"learning_rate": 9.36051282051282e-06,
"loss": 0.0427,
"step": 1750
},
{
"epoch": 4.23,
"learning_rate": 9.334871794871796e-06,
"loss": 0.0376,
"step": 1800
},
{
"epoch": 4.34,
"learning_rate": 9.30923076923077e-06,
"loss": 0.0478,
"step": 1850
},
{
"epoch": 4.46,
"learning_rate": 9.283589743589745e-06,
"loss": 0.0505,
"step": 1900
},
{
"epoch": 4.58,
"learning_rate": 9.257948717948719e-06,
"loss": 0.0397,
"step": 1950
},
{
"epoch": 4.69,
"learning_rate": 9.232307692307692e-06,
"loss": 0.0407,
"step": 2000
},
{
"epoch": 4.69,
"eval_loss": 0.2483939379453659,
"eval_runtime": 698.7199,
"eval_samples_per_second": 2.427,
"eval_steps_per_second": 0.607,
"eval_wer": 20.87667161961367,
"step": 2000
},
{
"epoch": 4.81,
"learning_rate": 9.206666666666668e-06,
"loss": 0.041,
"step": 2050
},
{
"epoch": 4.93,
"learning_rate": 9.181025641025641e-06,
"loss": 0.0508,
"step": 2100
},
{
"epoch": 5.05,
"learning_rate": 9.155384615384617e-06,
"loss": 0.0333,
"step": 2150
},
{
"epoch": 5.16,
"learning_rate": 9.12974358974359e-06,
"loss": 0.0272,
"step": 2200
},
{
"epoch": 5.28,
"learning_rate": 9.104102564102566e-06,
"loss": 0.0284,
"step": 2250
},
{
"epoch": 5.4,
"learning_rate": 9.07846153846154e-06,
"loss": 0.0233,
"step": 2300
},
{
"epoch": 5.52,
"learning_rate": 9.052820512820513e-06,
"loss": 0.0217,
"step": 2350
},
{
"epoch": 5.63,
"learning_rate": 9.027179487179488e-06,
"loss": 0.0304,
"step": 2400
},
{
"epoch": 5.75,
"learning_rate": 9.001538461538462e-06,
"loss": 0.0237,
"step": 2450
},
{
"epoch": 5.87,
"learning_rate": 8.975897435897437e-06,
"loss": 0.0267,
"step": 2500
},
{
"epoch": 5.99,
"learning_rate": 8.950256410256411e-06,
"loss": 0.0298,
"step": 2550
},
{
"epoch": 6.1,
"learning_rate": 8.924615384615385e-06,
"loss": 0.0264,
"step": 2600
},
{
"epoch": 6.22,
"learning_rate": 8.89897435897436e-06,
"loss": 0.022,
"step": 2650
},
{
"epoch": 6.34,
"learning_rate": 8.873333333333334e-06,
"loss": 0.0213,
"step": 2700
},
{
"epoch": 6.46,
"learning_rate": 8.847692307692309e-06,
"loss": 0.0195,
"step": 2750
},
{
"epoch": 6.57,
"learning_rate": 8.822051282051283e-06,
"loss": 0.0234,
"step": 2800
},
{
"epoch": 6.69,
"learning_rate": 8.796410256410258e-06,
"loss": 0.0214,
"step": 2850
},
{
"epoch": 6.81,
"learning_rate": 8.770769230769232e-06,
"loss": 0.0187,
"step": 2900
},
{
"epoch": 6.92,
"learning_rate": 8.745128205128205e-06,
"loss": 0.0226,
"step": 2950
},
{
"epoch": 7.04,
"learning_rate": 8.71948717948718e-06,
"loss": 0.0154,
"step": 3000
},
{
"epoch": 7.16,
"learning_rate": 8.693846153846154e-06,
"loss": 0.0187,
"step": 3050
},
{
"epoch": 7.28,
"learning_rate": 8.66820512820513e-06,
"loss": 0.0147,
"step": 3100
},
{
"epoch": 7.39,
"learning_rate": 8.642564102564103e-06,
"loss": 0.0104,
"step": 3150
},
{
"epoch": 7.51,
"learning_rate": 8.616923076923077e-06,
"loss": 0.0134,
"step": 3200
},
{
"epoch": 7.63,
"learning_rate": 8.591282051282052e-06,
"loss": 0.0178,
"step": 3250
},
{
"epoch": 7.75,
"learning_rate": 8.565641025641026e-06,
"loss": 0.0192,
"step": 3300
},
{
"epoch": 7.86,
"learning_rate": 8.540000000000001e-06,
"loss": 0.0169,
"step": 3350
},
{
"epoch": 7.98,
"learning_rate": 8.514358974358975e-06,
"loss": 0.0222,
"step": 3400
},
{
"epoch": 8.1,
"learning_rate": 8.48871794871795e-06,
"loss": 0.0132,
"step": 3450
},
{
"epoch": 8.22,
"learning_rate": 8.463076923076924e-06,
"loss": 0.0142,
"step": 3500
},
{
"epoch": 8.33,
"learning_rate": 8.437435897435898e-06,
"loss": 0.0119,
"step": 3550
},
{
"epoch": 8.45,
"learning_rate": 8.411794871794873e-06,
"loss": 0.0109,
"step": 3600
},
{
"epoch": 8.57,
"learning_rate": 8.386153846153847e-06,
"loss": 0.0132,
"step": 3650
},
{
"epoch": 8.69,
"learning_rate": 8.360512820512822e-06,
"loss": 0.0083,
"step": 3700
},
{
"epoch": 8.8,
"learning_rate": 8.334871794871796e-06,
"loss": 0.0106,
"step": 3750
},
{
"epoch": 8.92,
"learning_rate": 8.30923076923077e-06,
"loss": 0.0127,
"step": 3800
},
{
"epoch": 9.04,
"learning_rate": 8.283589743589745e-06,
"loss": 0.0115,
"step": 3850
},
{
"epoch": 9.15,
"learning_rate": 8.257948717948718e-06,
"loss": 0.0079,
"step": 3900
},
{
"epoch": 9.27,
"learning_rate": 8.232307692307694e-06,
"loss": 0.0078,
"step": 3950
},
{
"epoch": 9.39,
"learning_rate": 8.206666666666667e-06,
"loss": 0.0128,
"step": 4000
},
{
"epoch": 9.39,
"eval_loss": 0.2795361876487732,
"eval_runtime": 707.7086,
"eval_samples_per_second": 2.396,
"eval_steps_per_second": 0.599,
"eval_wer": 21.201708766716195,
"step": 4000
},
{
"epoch": 9.51,
"learning_rate": 8.181025641025642e-06,
"loss": 0.0122,
"step": 4050
},
{
"epoch": 9.62,
"learning_rate": 8.155384615384616e-06,
"loss": 0.0153,
"step": 4100
},
{
"epoch": 9.74,
"learning_rate": 8.13025641025641e-06,
"loss": 0.0136,
"step": 4150
},
{
"epoch": 9.86,
"learning_rate": 8.104615384615386e-06,
"loss": 0.0053,
"step": 4200
},
{
"epoch": 9.98,
"learning_rate": 8.07897435897436e-06,
"loss": 0.0122,
"step": 4250
},
{
"epoch": 10.09,
"learning_rate": 8.053333333333335e-06,
"loss": 0.0075,
"step": 4300
},
{
"epoch": 10.21,
"learning_rate": 8.027692307692308e-06,
"loss": 0.0056,
"step": 4350
},
{
"epoch": 10.33,
"learning_rate": 8.002051282051284e-06,
"loss": 0.0061,
"step": 4400
},
{
"epoch": 10.45,
"learning_rate": 7.976410256410257e-06,
"loss": 0.0039,
"step": 4450
},
{
"epoch": 10.56,
"learning_rate": 7.950769230769233e-06,
"loss": 0.0098,
"step": 4500
},
{
"epoch": 10.68,
"learning_rate": 7.925128205128205e-06,
"loss": 0.01,
"step": 4550
},
{
"epoch": 10.8,
"learning_rate": 7.89948717948718e-06,
"loss": 0.0062,
"step": 4600
},
{
"epoch": 10.92,
"learning_rate": 7.873846153846154e-06,
"loss": 0.009,
"step": 4650
},
{
"epoch": 11.03,
"learning_rate": 7.848205128205129e-06,
"loss": 0.0075,
"step": 4700
},
{
"epoch": 11.15,
"learning_rate": 7.822564102564103e-06,
"loss": 0.0066,
"step": 4750
},
{
"epoch": 11.27,
"learning_rate": 7.796923076923078e-06,
"loss": 0.0076,
"step": 4800
},
{
"epoch": 11.38,
"learning_rate": 7.771282051282052e-06,
"loss": 0.0087,
"step": 4850
},
{
"epoch": 11.5,
"learning_rate": 7.745641025641027e-06,
"loss": 0.0064,
"step": 4900
},
{
"epoch": 11.62,
"learning_rate": 7.72e-06,
"loss": 0.0049,
"step": 4950
},
{
"epoch": 11.74,
"learning_rate": 7.694358974358976e-06,
"loss": 0.0062,
"step": 5000
},
{
"epoch": 11.85,
"learning_rate": 7.66871794871795e-06,
"loss": 0.0092,
"step": 5050
},
{
"epoch": 11.97,
"learning_rate": 7.643076923076925e-06,
"loss": 0.0082,
"step": 5100
},
{
"epoch": 12.09,
"learning_rate": 7.617435897435898e-06,
"loss": 0.0071,
"step": 5150
},
{
"epoch": 12.21,
"learning_rate": 7.591794871794872e-06,
"loss": 0.0045,
"step": 5200
},
{
"epoch": 12.32,
"learning_rate": 7.566153846153847e-06,
"loss": 0.009,
"step": 5250
},
{
"epoch": 12.44,
"learning_rate": 7.540512820512821e-06,
"loss": 0.0079,
"step": 5300
},
{
"epoch": 12.56,
"learning_rate": 7.514871794871795e-06,
"loss": 0.0115,
"step": 5350
},
{
"epoch": 12.68,
"learning_rate": 7.489230769230769e-06,
"loss": 0.0073,
"step": 5400
},
{
"epoch": 12.79,
"learning_rate": 7.463589743589744e-06,
"loss": 0.0068,
"step": 5450
},
{
"epoch": 12.91,
"learning_rate": 7.437948717948718e-06,
"loss": 0.0094,
"step": 5500
},
{
"epoch": 13.03,
"learning_rate": 7.412307692307693e-06,
"loss": 0.0091,
"step": 5550
},
{
"epoch": 13.15,
"learning_rate": 7.386666666666667e-06,
"loss": 0.0045,
"step": 5600
},
{
"epoch": 13.26,
"learning_rate": 7.361025641025642e-06,
"loss": 0.004,
"step": 5650
},
{
"epoch": 13.38,
"learning_rate": 7.335384615384616e-06,
"loss": 0.0042,
"step": 5700
},
{
"epoch": 13.5,
"learning_rate": 7.309743589743591e-06,
"loss": 0.0067,
"step": 5750
},
{
"epoch": 13.62,
"learning_rate": 7.2841025641025645e-06,
"loss": 0.0077,
"step": 5800
},
{
"epoch": 13.73,
"learning_rate": 7.258461538461539e-06,
"loss": 0.014,
"step": 5850
},
{
"epoch": 13.85,
"learning_rate": 7.2328205128205135e-06,
"loss": 0.0076,
"step": 5900
},
{
"epoch": 13.97,
"learning_rate": 7.207179487179487e-06,
"loss": 0.0059,
"step": 5950
},
{
"epoch": 14.08,
"learning_rate": 7.181538461538462e-06,
"loss": 0.0041,
"step": 6000
},
{
"epoch": 14.08,
"eval_loss": 0.27444717288017273,
"eval_runtime": 695.0914,
"eval_samples_per_second": 2.44,
"eval_steps_per_second": 0.61,
"eval_wer": 19.13075780089153,
"step": 6000
},
{
"epoch": 14.2,
"learning_rate": 7.155897435897436e-06,
"loss": 0.0077,
"step": 6050
},
{
"epoch": 14.32,
"learning_rate": 7.130256410256411e-06,
"loss": 0.0055,
"step": 6100
},
{
"epoch": 14.44,
"learning_rate": 7.104615384615385e-06,
"loss": 0.0069,
"step": 6150
},
{
"epoch": 14.55,
"learning_rate": 7.07897435897436e-06,
"loss": 0.0049,
"step": 6200
},
{
"epoch": 14.67,
"learning_rate": 7.053333333333334e-06,
"loss": 0.006,
"step": 6250
},
{
"epoch": 14.79,
"learning_rate": 7.027692307692309e-06,
"loss": 0.0057,
"step": 6300
},
{
"epoch": 14.91,
"learning_rate": 7.002051282051283e-06,
"loss": 0.0066,
"step": 6350
},
{
"epoch": 15.02,
"learning_rate": 6.976410256410257e-06,
"loss": 0.0049,
"step": 6400
},
{
"epoch": 15.14,
"learning_rate": 6.950769230769231e-06,
"loss": 0.0034,
"step": 6450
},
{
"epoch": 15.26,
"learning_rate": 6.925128205128206e-06,
"loss": 0.004,
"step": 6500
},
{
"epoch": 15.38,
"learning_rate": 6.899487179487179e-06,
"loss": 0.0029,
"step": 6550
},
{
"epoch": 15.49,
"learning_rate": 6.873846153846154e-06,
"loss": 0.004,
"step": 6600
},
{
"epoch": 15.61,
"learning_rate": 6.848205128205128e-06,
"loss": 0.0044,
"step": 6650
},
{
"epoch": 15.73,
"learning_rate": 6.822564102564103e-06,
"loss": 0.0057,
"step": 6700
},
{
"epoch": 15.85,
"learning_rate": 6.796923076923077e-06,
"loss": 0.0052,
"step": 6750
},
{
"epoch": 15.96,
"learning_rate": 6.771282051282052e-06,
"loss": 0.0028,
"step": 6800
},
{
"epoch": 16.08,
"learning_rate": 6.745641025641026e-06,
"loss": 0.0049,
"step": 6850
},
{
"epoch": 16.2,
"learning_rate": 6.720000000000001e-06,
"loss": 0.0063,
"step": 6900
},
{
"epoch": 16.31,
"learning_rate": 6.694358974358975e-06,
"loss": 0.0049,
"step": 6950
},
{
"epoch": 16.43,
"learning_rate": 6.668717948717949e-06,
"loss": 0.005,
"step": 7000
},
{
"epoch": 16.55,
"learning_rate": 6.6430769230769235e-06,
"loss": 0.0055,
"step": 7050
},
{
"epoch": 16.67,
"learning_rate": 6.617435897435898e-06,
"loss": 0.0039,
"step": 7100
},
{
"epoch": 16.78,
"learning_rate": 6.5917948717948725e-06,
"loss": 0.0042,
"step": 7150
},
{
"epoch": 16.9,
"learning_rate": 6.566153846153846e-06,
"loss": 0.0041,
"step": 7200
},
{
"epoch": 17.02,
"learning_rate": 6.540512820512821e-06,
"loss": 0.0027,
"step": 7250
},
{
"epoch": 17.14,
"learning_rate": 6.514871794871795e-06,
"loss": 0.0031,
"step": 7300
},
{
"epoch": 17.25,
"learning_rate": 6.48923076923077e-06,
"loss": 0.0049,
"step": 7350
},
{
"epoch": 17.37,
"learning_rate": 6.463589743589744e-06,
"loss": 0.0046,
"step": 7400
},
{
"epoch": 17.49,
"learning_rate": 6.437948717948719e-06,
"loss": 0.0043,
"step": 7450
},
{
"epoch": 17.61,
"learning_rate": 6.412307692307693e-06,
"loss": 0.0041,
"step": 7500
},
{
"epoch": 17.72,
"learning_rate": 6.386666666666668e-06,
"loss": 0.0007,
"step": 7550
},
{
"epoch": 17.84,
"learning_rate": 6.361025641025641e-06,
"loss": 0.0049,
"step": 7600
},
{
"epoch": 17.96,
"learning_rate": 6.335384615384616e-06,
"loss": 0.0023,
"step": 7650
},
{
"epoch": 18.08,
"learning_rate": 6.30974358974359e-06,
"loss": 0.0039,
"step": 7700
},
{
"epoch": 18.19,
"learning_rate": 6.284102564102565e-06,
"loss": 0.0019,
"step": 7750
},
{
"epoch": 18.31,
"learning_rate": 6.258461538461538e-06,
"loss": 0.0045,
"step": 7800
},
{
"epoch": 18.43,
"learning_rate": 6.232820512820513e-06,
"loss": 0.0056,
"step": 7850
},
{
"epoch": 18.54,
"learning_rate": 6.207179487179487e-06,
"loss": 0.0044,
"step": 7900
},
{
"epoch": 18.66,
"learning_rate": 6.181538461538462e-06,
"loss": 0.003,
"step": 7950
},
{
"epoch": 18.78,
"learning_rate": 6.155897435897436e-06,
"loss": 0.0017,
"step": 8000
},
{
"epoch": 18.78,
"eval_loss": 0.2759210169315338,
"eval_runtime": 702.2209,
"eval_samples_per_second": 2.415,
"eval_steps_per_second": 0.604,
"eval_wer": 17.997771173848438,
"step": 8000
},
{
"epoch": 18.9,
"learning_rate": 6.130256410256411e-06,
"loss": 0.0053,
"step": 8050
},
{
"epoch": 19.01,
"learning_rate": 6.104615384615385e-06,
"loss": 0.0034,
"step": 8100
},
{
"epoch": 19.13,
"learning_rate": 6.07897435897436e-06,
"loss": 0.0032,
"step": 8150
},
{
"epoch": 19.25,
"learning_rate": 6.0533333333333335e-06,
"loss": 0.0025,
"step": 8200
},
{
"epoch": 19.37,
"learning_rate": 6.027692307692308e-06,
"loss": 0.005,
"step": 8250
},
{
"epoch": 19.48,
"learning_rate": 6.0020512820512825e-06,
"loss": 0.0027,
"step": 8300
},
{
"epoch": 19.6,
"learning_rate": 5.976410256410257e-06,
"loss": 0.0029,
"step": 8350
},
{
"epoch": 19.72,
"learning_rate": 5.950769230769231e-06,
"loss": 0.0041,
"step": 8400
},
{
"epoch": 19.84,
"learning_rate": 5.925128205128205e-06,
"loss": 0.0027,
"step": 8450
},
{
"epoch": 19.95,
"learning_rate": 5.89948717948718e-06,
"loss": 0.0045,
"step": 8500
},
{
"epoch": 20.07,
"learning_rate": 5.873846153846154e-06,
"loss": 0.005,
"step": 8550
},
{
"epoch": 20.19,
"learning_rate": 5.848205128205129e-06,
"loss": 0.0011,
"step": 8600
},
{
"epoch": 20.31,
"learning_rate": 5.822564102564103e-06,
"loss": 0.0021,
"step": 8650
},
{
"epoch": 20.42,
"learning_rate": 5.796923076923078e-06,
"loss": 0.0034,
"step": 8700
},
{
"epoch": 20.54,
"learning_rate": 5.771282051282052e-06,
"loss": 0.004,
"step": 8750
},
{
"epoch": 20.66,
"learning_rate": 5.745641025641027e-06,
"loss": 0.0022,
"step": 8800
},
{
"epoch": 20.77,
"learning_rate": 5.72e-06,
"loss": 0.0021,
"step": 8850
},
{
"epoch": 20.89,
"learning_rate": 5.694358974358975e-06,
"loss": 0.0029,
"step": 8900
},
{
"epoch": 21.01,
"learning_rate": 5.668717948717949e-06,
"loss": 0.0038,
"step": 8950
},
{
"epoch": 21.13,
"learning_rate": 5.643076923076923e-06,
"loss": 0.0019,
"step": 9000
},
{
"epoch": 21.24,
"learning_rate": 5.6174358974358974e-06,
"loss": 0.0023,
"step": 9050
},
{
"epoch": 21.36,
"learning_rate": 5.591794871794872e-06,
"loss": 0.0034,
"step": 9100
},
{
"epoch": 21.48,
"learning_rate": 5.566666666666667e-06,
"loss": 0.006,
"step": 9150
},
{
"epoch": 21.6,
"learning_rate": 5.5410256410256415e-06,
"loss": 0.0025,
"step": 9200
},
{
"epoch": 21.71,
"learning_rate": 5.515384615384616e-06,
"loss": 0.0019,
"step": 9250
},
{
"epoch": 21.83,
"learning_rate": 5.4897435897435905e-06,
"loss": 0.0013,
"step": 9300
},
{
"epoch": 21.95,
"learning_rate": 5.464102564102565e-06,
"loss": 0.0009,
"step": 9350
},
{
"epoch": 22.07,
"learning_rate": 5.4384615384615395e-06,
"loss": 0.0032,
"step": 9400
},
{
"epoch": 22.18,
"learning_rate": 5.412820512820514e-06,
"loss": 0.003,
"step": 9450
},
{
"epoch": 22.3,
"learning_rate": 5.387179487179488e-06,
"loss": 0.0068,
"step": 9500
},
{
"epoch": 22.42,
"learning_rate": 5.361538461538462e-06,
"loss": 0.0032,
"step": 9550
},
{
"epoch": 22.54,
"learning_rate": 5.335897435897436e-06,
"loss": 0.0065,
"step": 9600
},
{
"epoch": 22.65,
"learning_rate": 5.31025641025641e-06,
"loss": 0.0035,
"step": 9650
},
{
"epoch": 22.77,
"learning_rate": 5.284615384615385e-06,
"loss": 0.0014,
"step": 9700
},
{
"epoch": 22.89,
"learning_rate": 5.258974358974359e-06,
"loss": 0.0017,
"step": 9750
},
{
"epoch": 23.0,
"learning_rate": 5.233333333333334e-06,
"loss": 0.0019,
"step": 9800
},
{
"epoch": 23.12,
"learning_rate": 5.207692307692308e-06,
"loss": 0.0013,
"step": 9850
},
{
"epoch": 23.24,
"learning_rate": 5.182051282051283e-06,
"loss": 0.001,
"step": 9900
},
{
"epoch": 23.36,
"learning_rate": 5.156410256410257e-06,
"loss": 0.0003,
"step": 9950
},
{
"epoch": 23.47,
"learning_rate": 5.130769230769232e-06,
"loss": 0.0005,
"step": 10000
},
{
"epoch": 23.47,
"eval_loss": 0.2751367390155792,
"eval_runtime": 702.4615,
"eval_samples_per_second": 2.414,
"eval_steps_per_second": 0.604,
"eval_wer": 18.545690936106986,
"step": 10000
},
{
"epoch": 23.59,
"learning_rate": 5.105128205128206e-06,
"loss": 0.0018,
"step": 10050
},
{
"epoch": 23.71,
"learning_rate": 5.07948717948718e-06,
"loss": 0.0021,
"step": 10100
},
{
"epoch": 23.83,
"learning_rate": 5.053846153846154e-06,
"loss": 0.0022,
"step": 10150
},
{
"epoch": 23.94,
"learning_rate": 5.028205128205128e-06,
"loss": 0.0012,
"step": 10200
},
{
"epoch": 24.06,
"learning_rate": 5.0025641025641025e-06,
"loss": 0.0008,
"step": 10250
},
{
"epoch": 24.18,
"learning_rate": 4.976923076923078e-06,
"loss": 0.0009,
"step": 10300
},
{
"epoch": 24.3,
"learning_rate": 4.9512820512820515e-06,
"loss": 0.0011,
"step": 10350
},
{
"epoch": 24.41,
"learning_rate": 4.925641025641026e-06,
"loss": 0.0035,
"step": 10400
},
{
"epoch": 24.53,
"learning_rate": 4.9000000000000005e-06,
"loss": 0.002,
"step": 10450
},
{
"epoch": 24.65,
"learning_rate": 4.874358974358975e-06,
"loss": 0.0019,
"step": 10500
},
{
"epoch": 24.77,
"learning_rate": 4.8487179487179495e-06,
"loss": 0.0048,
"step": 10550
},
{
"epoch": 24.88,
"learning_rate": 4.823076923076924e-06,
"loss": 0.0031,
"step": 10600
},
{
"epoch": 25.0,
"learning_rate": 4.7974358974358985e-06,
"loss": 0.0003,
"step": 10650
},
{
"epoch": 25.12,
"learning_rate": 4.771794871794872e-06,
"loss": 0.0016,
"step": 10700
},
{
"epoch": 25.23,
"learning_rate": 4.746153846153847e-06,
"loss": 0.0009,
"step": 10750
},
{
"epoch": 25.35,
"learning_rate": 4.720512820512821e-06,
"loss": 0.0007,
"step": 10800
},
{
"epoch": 25.47,
"learning_rate": 4.694871794871796e-06,
"loss": 0.0011,
"step": 10850
},
{
"epoch": 25.59,
"learning_rate": 4.66923076923077e-06,
"loss": 0.0014,
"step": 10900
},
{
"epoch": 25.7,
"learning_rate": 4.643589743589745e-06,
"loss": 0.0016,
"step": 10950
},
{
"epoch": 25.82,
"learning_rate": 4.617948717948718e-06,
"loss": 0.0013,
"step": 11000
},
{
"epoch": 25.94,
"learning_rate": 4.592307692307693e-06,
"loss": 0.0019,
"step": 11050
},
{
"epoch": 26.06,
"learning_rate": 4.566666666666667e-06,
"loss": 0.0031,
"step": 11100
},
{
"epoch": 26.17,
"learning_rate": 4.541025641025642e-06,
"loss": 0.0033,
"step": 11150
},
{
"epoch": 26.29,
"learning_rate": 4.515384615384616e-06,
"loss": 0.0005,
"step": 11200
},
{
"epoch": 26.41,
"learning_rate": 4.489743589743591e-06,
"loss": 0.0025,
"step": 11250
},
{
"epoch": 26.53,
"learning_rate": 4.464102564102564e-06,
"loss": 0.0034,
"step": 11300
},
{
"epoch": 26.64,
"learning_rate": 4.438461538461539e-06,
"loss": 0.0011,
"step": 11350
},
{
"epoch": 26.76,
"learning_rate": 4.412820512820513e-06,
"loss": 0.0009,
"step": 11400
},
{
"epoch": 26.88,
"learning_rate": 4.387179487179488e-06,
"loss": 0.0009,
"step": 11450
},
{
"epoch": 27.0,
"learning_rate": 4.361538461538462e-06,
"loss": 0.0003,
"step": 11500
},
{
"epoch": 27.11,
"learning_rate": 4.335897435897437e-06,
"loss": 0.0011,
"step": 11550
},
{
"epoch": 27.23,
"learning_rate": 4.3102564102564105e-06,
"loss": 0.0014,
"step": 11600
},
{
"epoch": 27.35,
"learning_rate": 4.284615384615385e-06,
"loss": 0.0006,
"step": 11650
},
{
"epoch": 27.46,
"learning_rate": 4.2589743589743595e-06,
"loss": 0.0011,
"step": 11700
},
{
"epoch": 27.58,
"learning_rate": 4.233333333333334e-06,
"loss": 0.0022,
"step": 11750
},
{
"epoch": 27.7,
"learning_rate": 4.2076923076923085e-06,
"loss": 0.0015,
"step": 11800
},
{
"epoch": 27.82,
"learning_rate": 4.182051282051283e-06,
"loss": 0.0019,
"step": 11850
},
{
"epoch": 27.93,
"learning_rate": 4.156410256410257e-06,
"loss": 0.0004,
"step": 11900
},
{
"epoch": 28.05,
"learning_rate": 4.130769230769231e-06,
"loss": 0.001,
"step": 11950
},
{
"epoch": 28.17,
"learning_rate": 4.105128205128206e-06,
"loss": 0.0015,
"step": 12000
},
{
"epoch": 28.17,
"eval_loss": 0.2928332984447479,
"eval_runtime": 718.101,
"eval_samples_per_second": 2.362,
"eval_steps_per_second": 0.59,
"eval_wer": 19.20505200594354,
"step": 12000
},
{
"epoch": 28.29,
"learning_rate": 4.07948717948718e-06,
"loss": 0.0009,
"step": 12050
},
{
"epoch": 28.4,
"learning_rate": 4.053846153846155e-06,
"loss": 0.0028,
"step": 12100
},
{
"epoch": 28.52,
"learning_rate": 4.028205128205129e-06,
"loss": 0.0023,
"step": 12150
},
{
"epoch": 28.64,
"learning_rate": 4.002564102564103e-06,
"loss": 0.0016,
"step": 12200
},
{
"epoch": 28.76,
"learning_rate": 3.976923076923077e-06,
"loss": 0.0001,
"step": 12250
},
{
"epoch": 28.87,
"learning_rate": 3.951282051282052e-06,
"loss": 0.0011,
"step": 12300
},
{
"epoch": 28.99,
"learning_rate": 3.925641025641026e-06,
"loss": 0.0016,
"step": 12350
},
{
"epoch": 29.11,
"learning_rate": 3.900000000000001e-06,
"loss": 0.0009,
"step": 12400
},
{
"epoch": 29.23,
"learning_rate": 3.874358974358975e-06,
"loss": 0.0003,
"step": 12450
},
{
"epoch": 29.34,
"learning_rate": 3.848717948717949e-06,
"loss": 0.0004,
"step": 12500
},
{
"epoch": 29.46,
"learning_rate": 3.823076923076923e-06,
"loss": 0.0017,
"step": 12550
},
{
"epoch": 29.58,
"learning_rate": 3.7974358974358975e-06,
"loss": 0.0006,
"step": 12600
},
{
"epoch": 29.69,
"learning_rate": 3.771794871794872e-06,
"loss": 0.001,
"step": 12650
},
{
"epoch": 29.81,
"learning_rate": 3.7461538461538465e-06,
"loss": 0.0006,
"step": 12700
},
{
"epoch": 29.93,
"learning_rate": 3.720512820512821e-06,
"loss": 0.0017,
"step": 12750
},
{
"epoch": 30.05,
"learning_rate": 3.694871794871795e-06,
"loss": 0.0002,
"step": 12800
},
{
"epoch": 30.16,
"learning_rate": 3.6692307692307695e-06,
"loss": 0.002,
"step": 12850
},
{
"epoch": 30.28,
"learning_rate": 3.6435897435897436e-06,
"loss": 0.0004,
"step": 12900
},
{
"epoch": 30.4,
"learning_rate": 3.617948717948718e-06,
"loss": 0.0014,
"step": 12950
},
{
"epoch": 30.52,
"learning_rate": 3.5923076923076926e-06,
"loss": 0.0005,
"step": 13000
},
{
"epoch": 30.63,
"learning_rate": 3.566666666666667e-06,
"loss": 0.0019,
"step": 13050
},
{
"epoch": 30.75,
"learning_rate": 3.541025641025641e-06,
"loss": 0.0005,
"step": 13100
},
{
"epoch": 30.87,
"learning_rate": 3.5153846153846157e-06,
"loss": 0.0019,
"step": 13150
},
{
"epoch": 30.99,
"learning_rate": 3.4897435897435897e-06,
"loss": 0.001,
"step": 13200
},
{
"epoch": 31.1,
"learning_rate": 3.4641025641025642e-06,
"loss": 0.0002,
"step": 13250
},
{
"epoch": 31.22,
"learning_rate": 3.4384615384615387e-06,
"loss": 0.0019,
"step": 13300
},
{
"epoch": 31.34,
"learning_rate": 3.4128205128205132e-06,
"loss": 0.0008,
"step": 13350
},
{
"epoch": 31.46,
"learning_rate": 3.3871794871794873e-06,
"loss": 0.0013,
"step": 13400
},
{
"epoch": 31.57,
"learning_rate": 3.361538461538462e-06,
"loss": 0.0002,
"step": 13450
},
{
"epoch": 31.69,
"learning_rate": 3.3358974358974363e-06,
"loss": 0.001,
"step": 13500
},
{
"epoch": 31.81,
"learning_rate": 3.3102564102564104e-06,
"loss": 0.0031,
"step": 13550
},
{
"epoch": 31.92,
"learning_rate": 3.284615384615385e-06,
"loss": 0.0003,
"step": 13600
},
{
"epoch": 32.04,
"learning_rate": 3.2594871794871795e-06,
"loss": 0.0014,
"step": 13650
},
{
"epoch": 32.16,
"learning_rate": 3.233846153846154e-06,
"loss": 0.0001,
"step": 13700
},
{
"epoch": 32.28,
"learning_rate": 3.2082051282051285e-06,
"loss": 0.0003,
"step": 13750
},
{
"epoch": 32.39,
"learning_rate": 3.182564102564103e-06,
"loss": 0.0006,
"step": 13800
},
{
"epoch": 32.51,
"learning_rate": 3.1569230769230775e-06,
"loss": 0.001,
"step": 13850
},
{
"epoch": 32.63,
"learning_rate": 3.131282051282051e-06,
"loss": 0.001,
"step": 13900
},
{
"epoch": 32.75,
"learning_rate": 3.1056410256410257e-06,
"loss": 0.0013,
"step": 13950
},
{
"epoch": 32.86,
"learning_rate": 3.08e-06,
"loss": 0.0004,
"step": 14000
},
{
"epoch": 32.86,
"eval_loss": 0.2818757891654968,
"eval_runtime": 695.7492,
"eval_samples_per_second": 2.438,
"eval_steps_per_second": 0.609,
"eval_wer": 18.285661218424963,
"step": 14000
},
{
"epoch": 32.98,
"learning_rate": 3.0543589743589747e-06,
"loss": 0.0013,
"step": 14050
},
{
"epoch": 33.1,
"learning_rate": 3.028717948717949e-06,
"loss": 0.0004,
"step": 14100
},
{
"epoch": 33.22,
"learning_rate": 3.0030769230769236e-06,
"loss": 0.0006,
"step": 14150
},
{
"epoch": 33.33,
"learning_rate": 2.9774358974358973e-06,
"loss": 0.0006,
"step": 14200
},
{
"epoch": 33.45,
"learning_rate": 2.951794871794872e-06,
"loss": 0.0008,
"step": 14250
},
{
"epoch": 33.57,
"learning_rate": 2.9261538461538463e-06,
"loss": 0.0003,
"step": 14300
},
{
"epoch": 33.69,
"learning_rate": 2.9005128205128208e-06,
"loss": 0.0003,
"step": 14350
},
{
"epoch": 33.8,
"learning_rate": 2.8748717948717953e-06,
"loss": 0.0002,
"step": 14400
},
{
"epoch": 33.92,
"learning_rate": 2.8492307692307698e-06,
"loss": 0.001,
"step": 14450
},
{
"epoch": 34.04,
"learning_rate": 2.8235897435897434e-06,
"loss": 0.0003,
"step": 14500
},
{
"epoch": 34.15,
"learning_rate": 2.797948717948718e-06,
"loss": 0.0029,
"step": 14550
},
{
"epoch": 34.27,
"learning_rate": 2.7723076923076924e-06,
"loss": 0.0009,
"step": 14600
},
{
"epoch": 34.39,
"learning_rate": 2.746666666666667e-06,
"loss": 0.0012,
"step": 14650
},
{
"epoch": 34.51,
"learning_rate": 2.7210256410256414e-06,
"loss": 0.0014,
"step": 14700
},
{
"epoch": 34.62,
"learning_rate": 2.695384615384616e-06,
"loss": 0.0001,
"step": 14750
},
{
"epoch": 34.74,
"learning_rate": 2.6697435897435896e-06,
"loss": 0.0003,
"step": 14800
},
{
"epoch": 34.86,
"learning_rate": 2.644102564102564e-06,
"loss": 0.0009,
"step": 14850
},
{
"epoch": 34.98,
"learning_rate": 2.6184615384615385e-06,
"loss": 0.0001,
"step": 14900
},
{
"epoch": 35.09,
"learning_rate": 2.592820512820513e-06,
"loss": 0.0003,
"step": 14950
},
{
"epoch": 35.21,
"learning_rate": 2.5671794871794875e-06,
"loss": 0.0001,
"step": 15000
},
{
"epoch": 35.33,
"learning_rate": 2.541538461538462e-06,
"loss": 0.0003,
"step": 15050
},
{
"epoch": 35.45,
"learning_rate": 2.5158974358974357e-06,
"loss": 0.0012,
"step": 15100
},
{
"epoch": 35.56,
"learning_rate": 2.4902564102564106e-06,
"loss": 0.0017,
"step": 15150
},
{
"epoch": 35.68,
"learning_rate": 2.4646153846153847e-06,
"loss": 0.0006,
"step": 15200
},
{
"epoch": 35.8,
"learning_rate": 2.438974358974359e-06,
"loss": 0.001,
"step": 15250
},
{
"epoch": 35.92,
"learning_rate": 2.4133333333333337e-06,
"loss": 0.0013,
"step": 15300
},
{
"epoch": 36.03,
"learning_rate": 2.3876923076923077e-06,
"loss": 0.0001,
"step": 15350
},
{
"epoch": 36.15,
"learning_rate": 2.3620512820512822e-06,
"loss": 0.0015,
"step": 15400
},
{
"epoch": 36.27,
"learning_rate": 2.3364102564102567e-06,
"loss": 0.0001,
"step": 15450
},
{
"epoch": 36.38,
"learning_rate": 2.310769230769231e-06,
"loss": 0.0001,
"step": 15500
},
{
"epoch": 36.5,
"learning_rate": 2.2851282051282053e-06,
"loss": 0.0003,
"step": 15550
},
{
"epoch": 36.62,
"learning_rate": 2.25948717948718e-06,
"loss": 0.0009,
"step": 15600
},
{
"epoch": 36.74,
"learning_rate": 2.233846153846154e-06,
"loss": 0.0001,
"step": 15650
},
{
"epoch": 36.85,
"learning_rate": 2.2082051282051284e-06,
"loss": 0.0001,
"step": 15700
},
{
"epoch": 36.97,
"learning_rate": 2.182564102564103e-06,
"loss": 0.002,
"step": 15750
},
{
"epoch": 37.09,
"learning_rate": 2.156923076923077e-06,
"loss": 0.0009,
"step": 15800
},
{
"epoch": 37.21,
"learning_rate": 2.1312820512820514e-06,
"loss": 0.0005,
"step": 15850
},
{
"epoch": 37.32,
"learning_rate": 2.105641025641026e-06,
"loss": 0.0004,
"step": 15900
},
{
"epoch": 37.44,
"learning_rate": 2.08e-06,
"loss": 0.0001,
"step": 15950
},
{
"epoch": 37.56,
"learning_rate": 2.0543589743589745e-06,
"loss": 0.0002,
"step": 16000
},
{
"epoch": 37.56,
"eval_loss": 0.28313764929771423,
"eval_runtime": 695.718,
"eval_samples_per_second": 2.438,
"eval_steps_per_second": 0.609,
"eval_wer": 17.72845468053492,
"step": 16000
},
{
"epoch": 37.68,
"learning_rate": 2.028717948717949e-06,
"loss": 0.0002,
"step": 16050
},
{
"epoch": 37.79,
"learning_rate": 2.003076923076923e-06,
"loss": 0.0006,
"step": 16100
},
{
"epoch": 37.91,
"learning_rate": 1.9774358974358976e-06,
"loss": 0.0001,
"step": 16150
},
{
"epoch": 38.03,
"learning_rate": 1.951794871794872e-06,
"loss": 0.0002,
"step": 16200
},
{
"epoch": 38.15,
"learning_rate": 1.926153846153846e-06,
"loss": 0.0,
"step": 16250
},
{
"epoch": 38.26,
"learning_rate": 1.9005128205128206e-06,
"loss": 0.0,
"step": 16300
},
{
"epoch": 38.38,
"learning_rate": 1.8748717948717951e-06,
"loss": 0.0001,
"step": 16350
},
{
"epoch": 38.5,
"learning_rate": 1.8492307692307692e-06,
"loss": 0.0,
"step": 16400
},
{
"epoch": 38.62,
"learning_rate": 1.8235897435897437e-06,
"loss": 0.0,
"step": 16450
},
{
"epoch": 38.73,
"learning_rate": 1.7979487179487182e-06,
"loss": 0.0,
"step": 16500
},
{
"epoch": 38.85,
"learning_rate": 1.7723076923076922e-06,
"loss": 0.0,
"step": 16550
},
{
"epoch": 38.97,
"learning_rate": 1.7466666666666667e-06,
"loss": 0.0002,
"step": 16600
},
{
"epoch": 39.08,
"learning_rate": 1.7210256410256412e-06,
"loss": 0.0006,
"step": 16650
},
{
"epoch": 39.2,
"learning_rate": 1.6953846153846153e-06,
"loss": 0.0,
"step": 16700
},
{
"epoch": 39.32,
"learning_rate": 1.6697435897435898e-06,
"loss": 0.0003,
"step": 16750
},
{
"epoch": 39.44,
"learning_rate": 1.6441025641025643e-06,
"loss": 0.0,
"step": 16800
},
{
"epoch": 39.55,
"learning_rate": 1.6184615384615384e-06,
"loss": 0.0007,
"step": 16850
},
{
"epoch": 39.67,
"learning_rate": 1.5928205128205129e-06,
"loss": 0.0,
"step": 16900
},
{
"epoch": 39.79,
"learning_rate": 1.5676923076923078e-06,
"loss": 0.0008,
"step": 16950
},
{
"epoch": 39.91,
"learning_rate": 1.5420512820512822e-06,
"loss": 0.0004,
"step": 17000
},
{
"epoch": 40.02,
"learning_rate": 1.5164102564102565e-06,
"loss": 0.0011,
"step": 17050
},
{
"epoch": 40.14,
"learning_rate": 1.4907692307692308e-06,
"loss": 0.0003,
"step": 17100
},
{
"epoch": 40.26,
"learning_rate": 1.4651282051282053e-06,
"loss": 0.0013,
"step": 17150
},
{
"epoch": 40.38,
"learning_rate": 1.4394871794871796e-06,
"loss": 0.0003,
"step": 17200
},
{
"epoch": 40.49,
"learning_rate": 1.4138461538461539e-06,
"loss": 0.0001,
"step": 17250
},
{
"epoch": 40.61,
"learning_rate": 1.3882051282051284e-06,
"loss": 0.0004,
"step": 17300
},
{
"epoch": 40.73,
"learning_rate": 1.3625641025641027e-06,
"loss": 0.0005,
"step": 17350
},
{
"epoch": 40.85,
"learning_rate": 1.336923076923077e-06,
"loss": 0.001,
"step": 17400
},
{
"epoch": 40.96,
"learning_rate": 1.3112820512820514e-06,
"loss": 0.0006,
"step": 17450
},
{
"epoch": 41.08,
"learning_rate": 1.2856410256410257e-06,
"loss": 0.0001,
"step": 17500
},
{
"epoch": 41.2,
"learning_rate": 1.26e-06,
"loss": 0.0003,
"step": 17550
},
{
"epoch": 41.31,
"learning_rate": 1.2343589743589745e-06,
"loss": 0.0,
"step": 17600
},
{
"epoch": 41.43,
"learning_rate": 1.2087179487179488e-06,
"loss": 0.0001,
"step": 17650
},
{
"epoch": 41.55,
"learning_rate": 1.1830769230769233e-06,
"loss": 0.0003,
"step": 17700
},
{
"epoch": 41.67,
"learning_rate": 1.1574358974358976e-06,
"loss": 0.0003,
"step": 17750
},
{
"epoch": 41.78,
"learning_rate": 1.1317948717948719e-06,
"loss": 0.0001,
"step": 17800
},
{
"epoch": 41.9,
"learning_rate": 1.1061538461538463e-06,
"loss": 0.0,
"step": 17850
},
{
"epoch": 42.02,
"learning_rate": 1.0805128205128206e-06,
"loss": 0.0,
"step": 17900
},
{
"epoch": 42.14,
"learning_rate": 1.054871794871795e-06,
"loss": 0.0,
"step": 17950
},
{
"epoch": 42.25,
"learning_rate": 1.0292307692307694e-06,
"loss": 0.0007,
"step": 18000
},
{
"epoch": 42.25,
"eval_loss": 0.27763035893440247,
"eval_runtime": 699.0165,
"eval_samples_per_second": 2.426,
"eval_steps_per_second": 0.607,
"eval_wer": 17.83989598811293,
"step": 18000
},
{
"epoch": 42.37,
"learning_rate": 1.0035897435897437e-06,
"loss": 0.0008,
"step": 18050
},
{
"epoch": 42.49,
"learning_rate": 9.77948717948718e-07,
"loss": 0.0016,
"step": 18100
},
{
"epoch": 42.61,
"learning_rate": 9.523076923076924e-07,
"loss": 0.0001,
"step": 18150
},
{
"epoch": 42.72,
"learning_rate": 9.266666666666667e-07,
"loss": 0.0,
"step": 18200
},
{
"epoch": 42.84,
"learning_rate": 9.010256410256411e-07,
"loss": 0.0001,
"step": 18250
},
{
"epoch": 42.96,
"learning_rate": 8.753846153846154e-07,
"loss": 0.0,
"step": 18300
},
{
"epoch": 43.08,
"learning_rate": 8.497435897435897e-07,
"loss": 0.0014,
"step": 18350
},
{
"epoch": 43.19,
"learning_rate": 8.241025641025642e-07,
"loss": 0.0005,
"step": 18400
},
{
"epoch": 43.31,
"learning_rate": 7.984615384615385e-07,
"loss": 0.0,
"step": 18450
},
{
"epoch": 43.43,
"learning_rate": 7.728205128205128e-07,
"loss": 0.0001,
"step": 18500
},
{
"epoch": 43.54,
"learning_rate": 7.471794871794873e-07,
"loss": 0.0,
"step": 18550
},
{
"epoch": 43.66,
"learning_rate": 7.215384615384616e-07,
"loss": 0.0,
"step": 18600
},
{
"epoch": 43.78,
"learning_rate": 6.958974358974358e-07,
"loss": 0.0001,
"step": 18650
},
{
"epoch": 43.9,
"learning_rate": 6.702564102564103e-07,
"loss": 0.0,
"step": 18700
},
{
"epoch": 44.01,
"learning_rate": 6.446153846153846e-07,
"loss": 0.0,
"step": 18750
},
{
"epoch": 44.13,
"learning_rate": 6.18974358974359e-07,
"loss": 0.0,
"step": 18800
},
{
"epoch": 44.25,
"learning_rate": 5.933333333333334e-07,
"loss": 0.0001,
"step": 18850
},
{
"epoch": 44.37,
"learning_rate": 5.676923076923077e-07,
"loss": 0.0,
"step": 18900
},
{
"epoch": 44.48,
"learning_rate": 5.420512820512821e-07,
"loss": 0.0004,
"step": 18950
},
{
"epoch": 44.6,
"learning_rate": 5.164102564102565e-07,
"loss": 0.0,
"step": 19000
},
{
"epoch": 44.72,
"learning_rate": 4.907692307692308e-07,
"loss": 0.0001,
"step": 19050
},
{
"epoch": 44.84,
"learning_rate": 4.6512820512820514e-07,
"loss": 0.0,
"step": 19100
},
{
"epoch": 44.95,
"learning_rate": 4.3948717948717953e-07,
"loss": 0.0,
"step": 19150
},
{
"epoch": 45.07,
"learning_rate": 4.138461538461539e-07,
"loss": 0.0,
"step": 19200
},
{
"epoch": 45.19,
"learning_rate": 3.882051282051282e-07,
"loss": 0.0001,
"step": 19250
},
{
"epoch": 45.31,
"learning_rate": 3.625641025641026e-07,
"loss": 0.0015,
"step": 19300
},
{
"epoch": 45.42,
"learning_rate": 3.36923076923077e-07,
"loss": 0.0,
"step": 19350
},
{
"epoch": 45.54,
"learning_rate": 3.112820512820513e-07,
"loss": 0.0,
"step": 19400
},
{
"epoch": 45.66,
"learning_rate": 2.861538461538462e-07,
"loss": 0.0004,
"step": 19450
},
{
"epoch": 45.77,
"learning_rate": 2.6051282051282054e-07,
"loss": 0.0006,
"step": 19500
},
{
"epoch": 45.89,
"learning_rate": 2.348717948717949e-07,
"loss": 0.0,
"step": 19550
},
{
"epoch": 46.01,
"learning_rate": 2.0923076923076924e-07,
"loss": 0.0,
"step": 19600
},
{
"epoch": 46.13,
"learning_rate": 1.8358974358974358e-07,
"loss": 0.0003,
"step": 19650
},
{
"epoch": 46.24,
"learning_rate": 1.5794871794871797e-07,
"loss": 0.0,
"step": 19700
},
{
"epoch": 46.36,
"learning_rate": 1.323076923076923e-07,
"loss": 0.0,
"step": 19750
},
{
"epoch": 46.48,
"learning_rate": 1.0666666666666667e-07,
"loss": 0.0,
"step": 19800
},
{
"epoch": 46.6,
"learning_rate": 8.102564102564103e-08,
"loss": 0.0,
"step": 19850
},
{
"epoch": 46.71,
"learning_rate": 5.538461538461538e-08,
"loss": 0.0,
"step": 19900
},
{
"epoch": 46.83,
"learning_rate": 2.9743589743589746e-08,
"loss": 0.0,
"step": 19950
},
{
"epoch": 46.95,
"learning_rate": 4.102564102564102e-09,
"loss": 0.0,
"step": 20000
},
{
"epoch": 46.95,
"eval_loss": 0.2792465388774872,
"eval_runtime": 701.495,
"eval_samples_per_second": 2.418,
"eval_steps_per_second": 0.604,
"eval_wer": 17.09695393759287,
"step": 20000
},
{
"epoch": 46.95,
"step": 20000,
"total_flos": 1.6301509824872448e+20,
"train_loss": 0.029123614896199433,
"train_runtime": 43309.4036,
"train_samples_per_second": 3.694,
"train_steps_per_second": 0.462
}
],
"max_steps": 20000,
"num_train_epochs": 47,
"total_flos": 1.6301509824872448e+20,
"trial_name": null,
"trial_params": null
}