{ "best_metric": 11.839280651302198, "best_model_checkpoint": "./checkpoint-17000", "epoch": 97.0009, "eval_steps": 1000, "global_step": 20000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 4.0000000000000003e-07, "loss": 1.7621, "step": 25 }, { "epoch": 0.0, "learning_rate": 9.000000000000001e-07, "loss": 0.9336, "step": 50 }, { "epoch": 0.0, "learning_rate": 1.4000000000000001e-06, "loss": 0.7076, "step": 75 }, { "epoch": 0.01, "learning_rate": 1.9000000000000002e-06, "loss": 0.5828, "step": 100 }, { "epoch": 0.01, "learning_rate": 2.4000000000000003e-06, "loss": 0.5088, "step": 125 }, { "epoch": 0.01, "learning_rate": 2.9e-06, "loss": 0.4454, "step": 150 }, { "epoch": 0.01, "learning_rate": 3.4000000000000005e-06, "loss": 0.4004, "step": 175 }, { "epoch": 0.01, "learning_rate": 3.900000000000001e-06, "loss": 0.3607, "step": 200 }, { "epoch": 1.0, "learning_rate": 4.4e-06, "loss": 0.37, "step": 225 }, { "epoch": 1.0, "learning_rate": 4.9000000000000005e-06, "loss": 0.3235, "step": 250 }, { "epoch": 1.0, "learning_rate": 5.400000000000001e-06, "loss": 0.3303, "step": 275 }, { "epoch": 1.0, "learning_rate": 5.9e-06, "loss": 0.2879, "step": 300 }, { "epoch": 1.01, "learning_rate": 6.4000000000000006e-06, "loss": 0.2708, "step": 325 }, { "epoch": 1.01, "learning_rate": 6.9e-06, "loss": 0.2425, "step": 350 }, { "epoch": 1.01, "learning_rate": 7.4e-06, "loss": 0.2216, "step": 375 }, { "epoch": 1.01, "learning_rate": 7.9e-06, "loss": 0.1945, "step": 400 }, { "epoch": 2.0, "learning_rate": 8.400000000000001e-06, "loss": 0.2075, "step": 425 }, { "epoch": 2.0, "learning_rate": 8.900000000000001e-06, "loss": 0.189, "step": 450 }, { "epoch": 2.0, "learning_rate": 9.4e-06, "loss": 0.1839, "step": 475 }, { "epoch": 2.0, "learning_rate": 9.9e-06, "loss": 0.1727, "step": 500 }, { "epoch": 2.01, "learning_rate": 9.98974358974359e-06, "loss": 0.156, "step": 525 }, { "epoch": 2.01, "learning_rate": 9.976923076923077e-06, "loss": 0.1346, "step": 550 }, { "epoch": 2.01, "learning_rate": 9.964102564102564e-06, "loss": 0.1279, "step": 575 }, { "epoch": 2.01, "learning_rate": 9.951282051282051e-06, "loss": 0.1066, "step": 600 }, { "epoch": 3.0, "learning_rate": 9.93846153846154e-06, "loss": 0.1018, "step": 625 }, { "epoch": 3.0, "learning_rate": 9.925641025641026e-06, "loss": 0.0932, "step": 650 }, { "epoch": 3.0, "learning_rate": 9.912820512820513e-06, "loss": 0.0893, "step": 675 }, { "epoch": 3.0, "learning_rate": 9.9e-06, "loss": 0.0842, "step": 700 }, { "epoch": 3.01, "learning_rate": 9.887179487179489e-06, "loss": 0.0719, "step": 725 }, { "epoch": 3.01, "learning_rate": 9.874358974358975e-06, "loss": 0.0646, "step": 750 }, { "epoch": 3.01, "learning_rate": 9.861538461538462e-06, "loss": 0.0621, "step": 775 }, { "epoch": 3.01, "learning_rate": 9.848717948717949e-06, "loss": 0.0523, "step": 800 }, { "epoch": 4.0, "learning_rate": 9.835897435897438e-06, "loss": 0.0487, "step": 825 }, { "epoch": 4.0, "learning_rate": 9.823076923076924e-06, "loss": 0.0457, "step": 850 }, { "epoch": 4.0, "learning_rate": 9.810256410256411e-06, "loss": 0.0466, "step": 875 }, { "epoch": 4.0, "learning_rate": 9.797435897435898e-06, "loss": 0.0481, "step": 900 }, { "epoch": 4.01, "learning_rate": 9.784615384615387e-06, "loss": 0.0375, "step": 925 }, { "epoch": 4.01, "learning_rate": 9.771794871794873e-06, "loss": 0.036, "step": 950 }, { "epoch": 4.01, "learning_rate": 9.75897435897436e-06, "loss": 0.0333, "step": 975 }, { "epoch": 4.01, "learning_rate": 9.746153846153847e-06, "loss": 0.0293, "step": 1000 }, { "epoch": 4.01, "eval_loss": 0.27323344349861145, "eval_runtime": 2267.1838, "eval_samples_per_second": 2.907, "eval_steps_per_second": 0.182, "eval_wer": 15.948398072015877, "step": 1000 }, { "epoch": 4.01, "learning_rate": 9.733333333333334e-06, "loss": 0.0258, "step": 1025 }, { "epoch": 5.0, "learning_rate": 9.720512820512822e-06, "loss": 0.0263, "step": 1050 }, { "epoch": 5.0, "learning_rate": 9.707692307692308e-06, "loss": 0.025, "step": 1075 }, { "epoch": 5.0, "learning_rate": 9.694871794871794e-06, "loss": 0.0278, "step": 1100 }, { "epoch": 5.0, "learning_rate": 9.682051282051283e-06, "loss": 0.0222, "step": 1125 }, { "epoch": 5.01, "learning_rate": 9.66923076923077e-06, "loss": 0.0236, "step": 1150 }, { "epoch": 5.01, "learning_rate": 9.656410256410257e-06, "loss": 0.0213, "step": 1175 }, { "epoch": 5.01, "learning_rate": 9.643589743589743e-06, "loss": 0.0187, "step": 1200 }, { "epoch": 5.01, "learning_rate": 9.630769230769232e-06, "loss": 0.0144, "step": 1225 }, { "epoch": 6.0, "learning_rate": 9.617948717948719e-06, "loss": 0.0159, "step": 1250 }, { "epoch": 6.0, "learning_rate": 9.605128205128206e-06, "loss": 0.0162, "step": 1275 }, { "epoch": 6.0, "learning_rate": 9.592307692307692e-06, "loss": 0.018, "step": 1300 }, { "epoch": 6.0, "learning_rate": 9.579487179487181e-06, "loss": 0.0151, "step": 1325 }, { "epoch": 6.01, "learning_rate": 9.566666666666668e-06, "loss": 0.0135, "step": 1350 }, { "epoch": 6.01, "learning_rate": 9.553846153846155e-06, "loss": 0.012, "step": 1375 }, { "epoch": 6.01, "learning_rate": 9.541025641025641e-06, "loss": 0.0108, "step": 1400 }, { "epoch": 6.01, "learning_rate": 9.52820512820513e-06, "loss": 0.0097, "step": 1425 }, { "epoch": 7.0, "learning_rate": 9.515384615384617e-06, "loss": 0.0118, "step": 1450 }, { "epoch": 7.0, "learning_rate": 9.502564102564103e-06, "loss": 0.0098, "step": 1475 }, { "epoch": 7.0, "learning_rate": 9.48974358974359e-06, "loss": 0.0095, "step": 1500 }, { "epoch": 7.0, "learning_rate": 9.476923076923079e-06, "loss": 0.0083, "step": 1525 }, { "epoch": 7.01, "learning_rate": 9.464102564102566e-06, "loss": 0.0097, "step": 1550 }, { "epoch": 7.01, "learning_rate": 9.451282051282052e-06, "loss": 0.0078, "step": 1575 }, { "epoch": 7.01, "learning_rate": 9.43846153846154e-06, "loss": 0.0078, "step": 1600 }, { "epoch": 7.01, "learning_rate": 9.425641025641026e-06, "loss": 0.008, "step": 1625 }, { "epoch": 8.0, "learning_rate": 9.412820512820515e-06, "loss": 0.0083, "step": 1650 }, { "epoch": 8.0, "learning_rate": 9.4e-06, "loss": 0.007, "step": 1675 }, { "epoch": 8.0, "learning_rate": 9.387179487179488e-06, "loss": 0.0067, "step": 1700 }, { "epoch": 8.0, "learning_rate": 9.374358974358975e-06, "loss": 0.0068, "step": 1725 }, { "epoch": 8.01, "learning_rate": 9.361538461538462e-06, "loss": 0.0085, "step": 1750 }, { "epoch": 8.01, "learning_rate": 9.348717948717949e-06, "loss": 0.0081, "step": 1775 }, { "epoch": 8.01, "learning_rate": 9.335897435897436e-06, "loss": 0.0067, "step": 1800 }, { "epoch": 8.01, "learning_rate": 9.323076923076924e-06, "loss": 0.0067, "step": 1825 }, { "epoch": 8.01, "learning_rate": 9.310256410256411e-06, "loss": 0.0065, "step": 1850 }, { "epoch": 9.0, "learning_rate": 9.297435897435898e-06, "loss": 0.0061, "step": 1875 }, { "epoch": 9.0, "learning_rate": 9.284615384615385e-06, "loss": 0.0055, "step": 1900 }, { "epoch": 9.0, "learning_rate": 9.271794871794873e-06, "loss": 0.007, "step": 1925 }, { "epoch": 9.0, "learning_rate": 9.25897435897436e-06, "loss": 0.0069, "step": 1950 }, { "epoch": 9.01, "learning_rate": 9.246153846153847e-06, "loss": 0.0068, "step": 1975 }, { "epoch": 9.01, "learning_rate": 9.233333333333334e-06, "loss": 0.0065, "step": 2000 }, { "epoch": 9.01, "eval_loss": 0.3050675690174103, "eval_runtime": 2263.8207, "eval_samples_per_second": 2.911, "eval_steps_per_second": 0.182, "eval_wer": 14.113572846206813, "step": 2000 }, { "epoch": 9.01, "learning_rate": 9.220512820512822e-06, "loss": 0.0067, "step": 2025 }, { "epoch": 9.01, "learning_rate": 9.207692307692309e-06, "loss": 0.0052, "step": 2050 }, { "epoch": 10.0, "learning_rate": 9.194871794871796e-06, "loss": 0.0057, "step": 2075 }, { "epoch": 10.0, "learning_rate": 9.182051282051283e-06, "loss": 0.0065, "step": 2100 }, { "epoch": 10.0, "learning_rate": 9.169230769230771e-06, "loss": 0.0078, "step": 2125 }, { "epoch": 10.0, "learning_rate": 9.156410256410258e-06, "loss": 0.007, "step": 2150 }, { "epoch": 10.01, "learning_rate": 9.143589743589745e-06, "loss": 0.0066, "step": 2175 }, { "epoch": 10.01, "learning_rate": 9.130769230769232e-06, "loss": 0.0069, "step": 2200 }, { "epoch": 10.01, "learning_rate": 9.117948717948718e-06, "loss": 0.0067, "step": 2225 }, { "epoch": 10.01, "learning_rate": 9.105128205128207e-06, "loss": 0.0044, "step": 2250 }, { "epoch": 11.0, "learning_rate": 9.092307692307692e-06, "loss": 0.0056, "step": 2275 }, { "epoch": 11.0, "learning_rate": 9.07948717948718e-06, "loss": 0.0072, "step": 2300 }, { "epoch": 11.0, "learning_rate": 9.066666666666667e-06, "loss": 0.0067, "step": 2325 }, { "epoch": 11.0, "learning_rate": 9.053846153846154e-06, "loss": 0.007, "step": 2350 }, { "epoch": 11.01, "learning_rate": 9.041025641025641e-06, "loss": 0.0072, "step": 2375 }, { "epoch": 11.01, "learning_rate": 9.028205128205128e-06, "loss": 0.0064, "step": 2400 }, { "epoch": 11.01, "learning_rate": 9.015384615384616e-06, "loss": 0.0065, "step": 2425 }, { "epoch": 11.01, "learning_rate": 9.002564102564103e-06, "loss": 0.0062, "step": 2450 }, { "epoch": 12.0, "learning_rate": 8.98974358974359e-06, "loss": 0.0065, "step": 2475 }, { "epoch": 12.0, "learning_rate": 8.976923076923077e-06, "loss": 0.0065, "step": 2500 }, { "epoch": 12.0, "learning_rate": 8.964102564102565e-06, "loss": 0.0052, "step": 2525 }, { "epoch": 12.0, "learning_rate": 8.951282051282052e-06, "loss": 0.0062, "step": 2550 }, { "epoch": 12.01, "learning_rate": 8.938461538461539e-06, "loss": 0.0051, "step": 2575 }, { "epoch": 12.01, "learning_rate": 8.925641025641026e-06, "loss": 0.0056, "step": 2600 }, { "epoch": 12.01, "learning_rate": 8.912820512820514e-06, "loss": 0.0046, "step": 2625 }, { "epoch": 12.01, "learning_rate": 8.900000000000001e-06, "loss": 0.0074, "step": 2650 }, { "epoch": 12.01, "learning_rate": 8.887179487179488e-06, "loss": 0.0055, "step": 2675 }, { "epoch": 13.0, "learning_rate": 8.874358974358975e-06, "loss": 0.0064, "step": 2700 }, { "epoch": 13.0, "learning_rate": 8.861538461538463e-06, "loss": 0.0052, "step": 2725 }, { "epoch": 13.0, "learning_rate": 8.84871794871795e-06, "loss": 0.007, "step": 2750 }, { "epoch": 13.0, "learning_rate": 8.835897435897437e-06, "loss": 0.0058, "step": 2775 }, { "epoch": 13.01, "learning_rate": 8.823076923076924e-06, "loss": 0.004, "step": 2800 }, { "epoch": 13.01, "learning_rate": 8.81025641025641e-06, "loss": 0.0041, "step": 2825 }, { "epoch": 13.01, "learning_rate": 8.7974358974359e-06, "loss": 0.0036, "step": 2850 }, { "epoch": 13.01, "learning_rate": 8.784615384615386e-06, "loss": 0.0039, "step": 2875 }, { "epoch": 14.0, "learning_rate": 8.771794871794873e-06, "loss": 0.0043, "step": 2900 }, { "epoch": 14.0, "learning_rate": 8.75897435897436e-06, "loss": 0.0035, "step": 2925 }, { "epoch": 14.0, "learning_rate": 8.746153846153846e-06, "loss": 0.005, "step": 2950 }, { "epoch": 14.0, "learning_rate": 8.733333333333333e-06, "loss": 0.0039, "step": 2975 }, { "epoch": 14.01, "learning_rate": 8.72051282051282e-06, "loss": 0.0033, "step": 3000 }, { "epoch": 14.01, "eval_loss": 0.3100810647010803, "eval_runtime": 2225.815, "eval_samples_per_second": 2.961, "eval_steps_per_second": 0.185, "eval_wer": 13.240714488233627, "step": 3000 }, { "epoch": 14.01, "learning_rate": 8.707692307692309e-06, "loss": 0.0024, "step": 3025 }, { "epoch": 14.01, "learning_rate": 8.694871794871795e-06, "loss": 0.0032, "step": 3050 }, { "epoch": 14.01, "learning_rate": 8.682051282051282e-06, "loss": 0.0031, "step": 3075 }, { "epoch": 15.0, "learning_rate": 8.669230769230769e-06, "loss": 0.0039, "step": 3100 }, { "epoch": 15.0, "learning_rate": 8.656410256410258e-06, "loss": 0.0031, "step": 3125 }, { "epoch": 15.0, "learning_rate": 8.643589743589744e-06, "loss": 0.0025, "step": 3150 }, { "epoch": 15.0, "learning_rate": 8.630769230769231e-06, "loss": 0.0036, "step": 3175 }, { "epoch": 15.01, "learning_rate": 8.617948717948718e-06, "loss": 0.0035, "step": 3200 }, { "epoch": 15.01, "learning_rate": 8.605128205128207e-06, "loss": 0.003, "step": 3225 }, { "epoch": 15.01, "learning_rate": 8.592307692307693e-06, "loss": 0.0027, "step": 3250 }, { "epoch": 15.01, "learning_rate": 8.57948717948718e-06, "loss": 0.0031, "step": 3275 }, { "epoch": 16.0, "learning_rate": 8.566666666666667e-06, "loss": 0.0029, "step": 3300 }, { "epoch": 16.0, "learning_rate": 8.553846153846156e-06, "loss": 0.0022, "step": 3325 }, { "epoch": 16.0, "learning_rate": 8.541025641025642e-06, "loss": 0.003, "step": 3350 }, { "epoch": 16.0, "learning_rate": 8.52820512820513e-06, "loss": 0.0035, "step": 3375 }, { "epoch": 16.01, "learning_rate": 8.515384615384616e-06, "loss": 0.0038, "step": 3400 }, { "epoch": 16.01, "learning_rate": 8.502564102564103e-06, "loss": 0.0025, "step": 3425 }, { "epoch": 16.01, "learning_rate": 8.489743589743591e-06, "loss": 0.0024, "step": 3450 }, { "epoch": 16.01, "learning_rate": 8.476923076923078e-06, "loss": 0.003, "step": 3475 }, { "epoch": 16.01, "learning_rate": 8.464102564102565e-06, "loss": 0.0035, "step": 3500 }, { "epoch": 17.0, "learning_rate": 8.451282051282052e-06, "loss": 0.0038, "step": 3525 }, { "epoch": 17.0, "learning_rate": 8.438461538461539e-06, "loss": 0.003, "step": 3550 }, { "epoch": 17.0, "learning_rate": 8.425641025641026e-06, "loss": 0.0031, "step": 3575 }, { "epoch": 17.0, "learning_rate": 8.412820512820512e-06, "loss": 0.0035, "step": 3600 }, { "epoch": 17.01, "learning_rate": 8.400000000000001e-06, "loss": 0.0033, "step": 3625 }, { "epoch": 17.01, "learning_rate": 8.387179487179488e-06, "loss": 0.0028, "step": 3650 }, { "epoch": 17.01, "learning_rate": 8.374358974358975e-06, "loss": 0.0022, "step": 3675 }, { "epoch": 17.01, "learning_rate": 8.361538461538461e-06, "loss": 0.0021, "step": 3700 }, { "epoch": 18.0, "learning_rate": 8.34871794871795e-06, "loss": 0.0032, "step": 3725 }, { "epoch": 18.0, "learning_rate": 8.335897435897437e-06, "loss": 0.0029, "step": 3750 }, { "epoch": 18.0, "learning_rate": 8.323076923076924e-06, "loss": 0.0041, "step": 3775 }, { "epoch": 18.0, "learning_rate": 8.31025641025641e-06, "loss": 0.0042, "step": 3800 }, { "epoch": 18.01, "learning_rate": 8.297435897435899e-06, "loss": 0.0036, "step": 3825 }, { "epoch": 18.01, "learning_rate": 8.284615384615386e-06, "loss": 0.0047, "step": 3850 }, { "epoch": 18.01, "learning_rate": 8.271794871794873e-06, "loss": 0.005, "step": 3875 }, { "epoch": 18.01, "learning_rate": 8.25897435897436e-06, "loss": 0.005, "step": 3900 }, { "epoch": 19.0, "learning_rate": 8.246153846153848e-06, "loss": 0.0039, "step": 3925 }, { "epoch": 19.0, "learning_rate": 8.233333333333335e-06, "loss": 0.0033, "step": 3950 }, { "epoch": 19.0, "learning_rate": 8.220512820512822e-06, "loss": 0.0035, "step": 3975 }, { "epoch": 19.0, "learning_rate": 8.207692307692308e-06, "loss": 0.0041, "step": 4000 }, { "epoch": 19.0, "eval_loss": 0.31360939145088196, "eval_runtime": 2287.1242, "eval_samples_per_second": 2.882, "eval_steps_per_second": 0.18, "eval_wer": 13.830045769370974, "step": 4000 }, { "epoch": 19.01, "learning_rate": 8.194871794871795e-06, "loss": 0.0049, "step": 4025 }, { "epoch": 19.01, "learning_rate": 8.182051282051284e-06, "loss": 0.0034, "step": 4050 }, { "epoch": 19.01, "learning_rate": 8.16923076923077e-06, "loss": 0.0047, "step": 4075 }, { "epoch": 19.01, "learning_rate": 8.156410256410257e-06, "loss": 0.0054, "step": 4100 }, { "epoch": 20.0, "learning_rate": 8.143589743589744e-06, "loss": 0.0054, "step": 4125 }, { "epoch": 20.0, "learning_rate": 8.130769230769231e-06, "loss": 0.0062, "step": 4150 }, { "epoch": 20.0, "learning_rate": 8.117948717948718e-06, "loss": 0.0046, "step": 4175 }, { "epoch": 20.0, "learning_rate": 8.105128205128205e-06, "loss": 0.0048, "step": 4200 }, { "epoch": 20.01, "learning_rate": 8.092307692307693e-06, "loss": 0.0038, "step": 4225 }, { "epoch": 20.01, "learning_rate": 8.07948717948718e-06, "loss": 0.0027, "step": 4250 }, { "epoch": 20.01, "learning_rate": 8.066666666666667e-06, "loss": 0.0034, "step": 4275 }, { "epoch": 20.01, "learning_rate": 8.053846153846154e-06, "loss": 0.0054, "step": 4300 }, { "epoch": 20.01, "learning_rate": 8.041025641025642e-06, "loss": 0.0038, "step": 4325 }, { "epoch": 21.0, "learning_rate": 8.028205128205129e-06, "loss": 0.0031, "step": 4350 }, { "epoch": 21.0, "learning_rate": 8.015384615384616e-06, "loss": 0.0043, "step": 4375 }, { "epoch": 21.0, "learning_rate": 8.002564102564103e-06, "loss": 0.0033, "step": 4400 }, { "epoch": 21.0, "learning_rate": 7.989743589743591e-06, "loss": 0.0038, "step": 4425 }, { "epoch": 21.01, "learning_rate": 7.976923076923078e-06, "loss": 0.0031, "step": 4450 }, { "epoch": 21.01, "learning_rate": 7.964102564102565e-06, "loss": 0.0032, "step": 4475 }, { "epoch": 21.01, "learning_rate": 7.951282051282052e-06, "loss": 0.0037, "step": 4500 }, { "epoch": 21.01, "learning_rate": 7.93846153846154e-06, "loss": 0.0027, "step": 4525 }, { "epoch": 22.0, "learning_rate": 7.925641025641027e-06, "loss": 0.0028, "step": 4550 }, { "epoch": 22.0, "learning_rate": 7.912820512820514e-06, "loss": 0.0025, "step": 4575 }, { "epoch": 22.0, "learning_rate": 7.9e-06, "loss": 0.0028, "step": 4600 }, { "epoch": 22.0, "learning_rate": 7.887179487179487e-06, "loss": 0.0019, "step": 4625 }, { "epoch": 22.01, "learning_rate": 7.874358974358976e-06, "loss": 0.0015, "step": 4650 }, { "epoch": 22.01, "learning_rate": 7.861538461538463e-06, "loss": 0.0026, "step": 4675 }, { "epoch": 22.01, "learning_rate": 7.84871794871795e-06, "loss": 0.0016, "step": 4700 }, { "epoch": 22.01, "learning_rate": 7.835897435897436e-06, "loss": 0.0016, "step": 4725 }, { "epoch": 23.0, "learning_rate": 7.823076923076923e-06, "loss": 0.0021, "step": 4750 }, { "epoch": 23.0, "learning_rate": 7.81025641025641e-06, "loss": 0.0019, "step": 4775 }, { "epoch": 23.0, "learning_rate": 7.797435897435897e-06, "loss": 0.0014, "step": 4800 }, { "epoch": 23.0, "learning_rate": 7.784615384615385e-06, "loss": 0.0016, "step": 4825 }, { "epoch": 23.01, "learning_rate": 7.771794871794872e-06, "loss": 0.0013, "step": 4850 }, { "epoch": 23.01, "learning_rate": 7.758974358974359e-06, "loss": 0.0022, "step": 4875 }, { "epoch": 23.01, "learning_rate": 7.746153846153846e-06, "loss": 0.0015, "step": 4900 }, { "epoch": 23.01, "learning_rate": 7.733333333333334e-06, "loss": 0.0012, "step": 4925 }, { "epoch": 24.0, "learning_rate": 7.720512820512821e-06, "loss": 0.0016, "step": 4950 }, { "epoch": 24.0, "learning_rate": 7.707692307692308e-06, "loss": 0.0013, "step": 4975 }, { "epoch": 24.0, "learning_rate": 7.694871794871795e-06, "loss": 0.0013, "step": 5000 }, { "epoch": 24.0, "eval_loss": 0.31794673204421997, "eval_runtime": 2265.7471, "eval_samples_per_second": 2.909, "eval_steps_per_second": 0.182, "eval_wer": 12.73644133014703, "step": 5000 }, { "epoch": 24.0, "learning_rate": 7.682051282051283e-06, "loss": 0.0014, "step": 5025 }, { "epoch": 24.01, "learning_rate": 7.66923076923077e-06, "loss": 0.0017, "step": 5050 }, { "epoch": 24.01, "learning_rate": 7.656410256410257e-06, "loss": 0.0012, "step": 5075 }, { "epoch": 24.01, "learning_rate": 7.643589743589744e-06, "loss": 0.0014, "step": 5100 }, { "epoch": 24.01, "learning_rate": 7.630769230769232e-06, "loss": 0.0022, "step": 5125 }, { "epoch": 24.01, "learning_rate": 7.617948717948719e-06, "loss": 0.0013, "step": 5150 }, { "epoch": 25.0, "learning_rate": 7.605128205128206e-06, "loss": 0.0025, "step": 5175 }, { "epoch": 25.0, "learning_rate": 7.592307692307693e-06, "loss": 0.0022, "step": 5200 }, { "epoch": 25.0, "learning_rate": 7.5794871794871805e-06, "loss": 0.0017, "step": 5225 }, { "epoch": 25.0, "learning_rate": 7.566666666666667e-06, "loss": 0.002, "step": 5250 }, { "epoch": 25.01, "learning_rate": 7.553846153846155e-06, "loss": 0.0025, "step": 5275 }, { "epoch": 25.01, "learning_rate": 7.541025641025642e-06, "loss": 0.0013, "step": 5300 }, { "epoch": 25.01, "learning_rate": 7.528205128205129e-06, "loss": 0.0019, "step": 5325 }, { "epoch": 25.01, "learning_rate": 7.5153846153846155e-06, "loss": 0.0017, "step": 5350 }, { "epoch": 26.0, "learning_rate": 7.502564102564102e-06, "loss": 0.0018, "step": 5375 }, { "epoch": 26.0, "learning_rate": 7.48974358974359e-06, "loss": 0.0015, "step": 5400 }, { "epoch": 26.0, "learning_rate": 7.476923076923077e-06, "loss": 0.0019, "step": 5425 }, { "epoch": 26.0, "learning_rate": 7.4641025641025645e-06, "loss": 0.0017, "step": 5450 }, { "epoch": 26.01, "learning_rate": 7.451282051282051e-06, "loss": 0.0018, "step": 5475 }, { "epoch": 26.01, "learning_rate": 7.438461538461539e-06, "loss": 0.0028, "step": 5500 }, { "epoch": 26.01, "learning_rate": 7.425641025641026e-06, "loss": 0.0026, "step": 5525 }, { "epoch": 26.01, "learning_rate": 7.4128205128205135e-06, "loss": 0.0017, "step": 5550 }, { "epoch": 27.0, "learning_rate": 7.4e-06, "loss": 0.0019, "step": 5575 }, { "epoch": 27.0, "learning_rate": 7.387179487179488e-06, "loss": 0.002, "step": 5600 }, { "epoch": 27.0, "learning_rate": 7.374358974358975e-06, "loss": 0.0027, "step": 5625 }, { "epoch": 27.0, "learning_rate": 7.3615384615384625e-06, "loss": 0.0025, "step": 5650 }, { "epoch": 27.01, "learning_rate": 7.348717948717949e-06, "loss": 0.0029, "step": 5675 }, { "epoch": 27.01, "learning_rate": 7.335897435897437e-06, "loss": 0.0032, "step": 5700 }, { "epoch": 27.01, "learning_rate": 7.323076923076924e-06, "loss": 0.0045, "step": 5725 }, { "epoch": 27.01, "learning_rate": 7.3102564102564115e-06, "loss": 0.004, "step": 5750 }, { "epoch": 28.0, "learning_rate": 7.297435897435898e-06, "loss": 0.0037, "step": 5775 }, { "epoch": 28.0, "learning_rate": 7.284615384615386e-06, "loss": 0.0027, "step": 5800 }, { "epoch": 28.0, "learning_rate": 7.271794871794873e-06, "loss": 0.0026, "step": 5825 }, { "epoch": 28.0, "learning_rate": 7.25897435897436e-06, "loss": 0.0033, "step": 5850 }, { "epoch": 28.01, "learning_rate": 7.246153846153847e-06, "loss": 0.0037, "step": 5875 }, { "epoch": 28.01, "learning_rate": 7.233333333333334e-06, "loss": 0.0036, "step": 5900 }, { "epoch": 28.01, "learning_rate": 7.220512820512822e-06, "loss": 0.0041, "step": 5925 }, { "epoch": 28.01, "learning_rate": 7.207692307692308e-06, "loss": 0.0053, "step": 5950 }, { "epoch": 29.0, "learning_rate": 7.1948717948717946e-06, "loss": 0.0036, "step": 5975 }, { "epoch": 29.0, "learning_rate": 7.182051282051282e-06, "loss": 0.0046, "step": 6000 }, { "epoch": 29.0, "eval_loss": 0.3209567964076996, "eval_runtime": 2302.6465, "eval_samples_per_second": 2.862, "eval_steps_per_second": 0.179, "eval_wer": 13.663979910081414, "step": 6000 }, { "epoch": 29.0, "learning_rate": 7.169230769230769e-06, "loss": 0.0036, "step": 6025 }, { "epoch": 29.0, "learning_rate": 7.156410256410257e-06, "loss": 0.0035, "step": 6050 }, { "epoch": 29.01, "learning_rate": 7.1435897435897436e-06, "loss": 0.0036, "step": 6075 }, { "epoch": 29.01, "learning_rate": 7.130769230769231e-06, "loss": 0.0046, "step": 6100 }, { "epoch": 29.01, "learning_rate": 7.117948717948718e-06, "loss": 0.0032, "step": 6125 }, { "epoch": 29.01, "learning_rate": 7.105128205128206e-06, "loss": 0.0028, "step": 6150 }, { "epoch": 29.01, "learning_rate": 7.0923076923076926e-06, "loss": 0.0038, "step": 6175 }, { "epoch": 30.0, "learning_rate": 7.07948717948718e-06, "loss": 0.0031, "step": 6200 }, { "epoch": 30.0, "learning_rate": 7.066666666666667e-06, "loss": 0.0035, "step": 6225 }, { "epoch": 30.0, "learning_rate": 7.053846153846155e-06, "loss": 0.0031, "step": 6250 }, { "epoch": 30.0, "learning_rate": 7.0410256410256415e-06, "loss": 0.0029, "step": 6275 }, { "epoch": 30.01, "learning_rate": 7.028205128205129e-06, "loss": 0.0026, "step": 6300 }, { "epoch": 30.01, "learning_rate": 7.015384615384616e-06, "loss": 0.0025, "step": 6325 }, { "epoch": 30.01, "learning_rate": 7.002564102564104e-06, "loss": 0.0024, "step": 6350 }, { "epoch": 30.01, "learning_rate": 6.9897435897435905e-06, "loss": 0.0019, "step": 6375 }, { "epoch": 31.0, "learning_rate": 6.976923076923078e-06, "loss": 0.0023, "step": 6400 }, { "epoch": 31.0, "learning_rate": 6.964102564102565e-06, "loss": 0.0031, "step": 6425 }, { "epoch": 31.0, "learning_rate": 6.951282051282052e-06, "loss": 0.0017, "step": 6450 }, { "epoch": 31.0, "learning_rate": 6.9384615384615395e-06, "loss": 0.0021, "step": 6475 }, { "epoch": 31.01, "learning_rate": 6.925641025641026e-06, "loss": 0.0015, "step": 6500 }, { "epoch": 31.01, "learning_rate": 6.912820512820514e-06, "loss": 0.0017, "step": 6525 }, { "epoch": 31.01, "learning_rate": 6.9e-06, "loss": 0.0013, "step": 6550 }, { "epoch": 31.01, "learning_rate": 6.887179487179488e-06, "loss": 0.0014, "step": 6575 }, { "epoch": 32.0, "learning_rate": 6.8743589743589745e-06, "loss": 0.0019, "step": 6600 }, { "epoch": 32.0, "learning_rate": 6.861538461538461e-06, "loss": 0.0016, "step": 6625 }, { "epoch": 32.0, "learning_rate": 6.848717948717949e-06, "loss": 0.001, "step": 6650 }, { "epoch": 32.0, "learning_rate": 6.835897435897436e-06, "loss": 0.0016, "step": 6675 }, { "epoch": 32.01, "learning_rate": 6.8230769230769235e-06, "loss": 0.0014, "step": 6700 }, { "epoch": 32.01, "learning_rate": 6.81025641025641e-06, "loss": 0.0013, "step": 6725 }, { "epoch": 32.01, "learning_rate": 6.797435897435898e-06, "loss": 0.0016, "step": 6750 }, { "epoch": 32.01, "learning_rate": 6.784615384615385e-06, "loss": 0.001, "step": 6775 }, { "epoch": 33.0, "learning_rate": 6.7717948717948725e-06, "loss": 0.0013, "step": 6800 }, { "epoch": 33.0, "learning_rate": 6.758974358974359e-06, "loss": 0.001, "step": 6825 }, { "epoch": 33.0, "learning_rate": 6.746153846153847e-06, "loss": 0.0015, "step": 6850 }, { "epoch": 33.0, "learning_rate": 6.733333333333334e-06, "loss": 0.0011, "step": 6875 }, { "epoch": 33.01, "learning_rate": 6.7205128205128215e-06, "loss": 0.0007, "step": 6900 }, { "epoch": 33.01, "learning_rate": 6.707692307692308e-06, "loss": 0.0014, "step": 6925 }, { "epoch": 33.01, "learning_rate": 6.694871794871796e-06, "loss": 0.0014, "step": 6950 }, { "epoch": 33.01, "learning_rate": 6.682051282051283e-06, "loss": 0.0015, "step": 6975 }, { "epoch": 33.01, "learning_rate": 6.6692307692307705e-06, "loss": 0.0015, "step": 7000 }, { "epoch": 33.01, "eval_loss": 0.3261638879776001, "eval_runtime": 2317.8239, "eval_samples_per_second": 2.844, "eval_steps_per_second": 0.178, "eval_wer": 12.80934829276196, "step": 7000 }, { "epoch": 34.0, "learning_rate": 6.656410256410257e-06, "loss": 0.0008, "step": 7025 }, { "epoch": 34.0, "learning_rate": 6.643589743589744e-06, "loss": 0.0006, "step": 7050 }, { "epoch": 34.0, "learning_rate": 6.630769230769232e-06, "loss": 0.0009, "step": 7075 }, { "epoch": 34.0, "learning_rate": 6.617948717948719e-06, "loss": 0.001, "step": 7100 }, { "epoch": 34.01, "learning_rate": 6.605128205128206e-06, "loss": 0.0008, "step": 7125 }, { "epoch": 34.01, "learning_rate": 6.592307692307692e-06, "loss": 0.0009, "step": 7150 }, { "epoch": 34.01, "learning_rate": 6.57948717948718e-06, "loss": 0.0015, "step": 7175 }, { "epoch": 34.01, "learning_rate": 6.566666666666667e-06, "loss": 0.0008, "step": 7200 }, { "epoch": 35.0, "learning_rate": 6.553846153846154e-06, "loss": 0.0005, "step": 7225 }, { "epoch": 35.0, "learning_rate": 6.541025641025641e-06, "loss": 0.0009, "step": 7250 }, { "epoch": 35.0, "learning_rate": 6.528205128205128e-06, "loss": 0.0008, "step": 7275 }, { "epoch": 35.0, "learning_rate": 6.515384615384616e-06, "loss": 0.0006, "step": 7300 }, { "epoch": 35.01, "learning_rate": 6.5025641025641026e-06, "loss": 0.001, "step": 7325 }, { "epoch": 35.01, "learning_rate": 6.48974358974359e-06, "loss": 0.0009, "step": 7350 }, { "epoch": 35.01, "learning_rate": 6.476923076923077e-06, "loss": 0.0005, "step": 7375 }, { "epoch": 35.01, "learning_rate": 6.464102564102565e-06, "loss": 0.0006, "step": 7400 }, { "epoch": 36.0, "learning_rate": 6.4512820512820516e-06, "loss": 0.0009, "step": 7425 }, { "epoch": 36.0, "learning_rate": 6.438461538461539e-06, "loss": 0.0012, "step": 7450 }, { "epoch": 36.0, "learning_rate": 6.425641025641026e-06, "loss": 0.0012, "step": 7475 }, { "epoch": 36.0, "learning_rate": 6.412820512820514e-06, "loss": 0.0013, "step": 7500 }, { "epoch": 36.01, "learning_rate": 6.4000000000000006e-06, "loss": 0.0014, "step": 7525 }, { "epoch": 36.01, "learning_rate": 6.387179487179488e-06, "loss": 0.0011, "step": 7550 }, { "epoch": 36.01, "learning_rate": 6.374358974358975e-06, "loss": 0.0021, "step": 7575 }, { "epoch": 36.01, "learning_rate": 6.361538461538463e-06, "loss": 0.001, "step": 7600 }, { "epoch": 37.0, "learning_rate": 6.3487179487179495e-06, "loss": 0.0016, "step": 7625 }, { "epoch": 37.0, "learning_rate": 6.335897435897436e-06, "loss": 0.0008, "step": 7650 }, { "epoch": 37.0, "learning_rate": 6.323076923076924e-06, "loss": 0.0014, "step": 7675 }, { "epoch": 37.0, "learning_rate": 6.310256410256411e-06, "loss": 0.0013, "step": 7700 }, { "epoch": 37.01, "learning_rate": 6.2974358974358985e-06, "loss": 0.0019, "step": 7725 }, { "epoch": 37.01, "learning_rate": 6.284615384615385e-06, "loss": 0.0017, "step": 7750 }, { "epoch": 37.01, "learning_rate": 6.271794871794872e-06, "loss": 0.002, "step": 7775 }, { "epoch": 37.01, "learning_rate": 6.258974358974359e-06, "loss": 0.0014, "step": 7800 }, { "epoch": 37.01, "learning_rate": 6.246153846153846e-06, "loss": 0.0015, "step": 7825 }, { "epoch": 38.0, "learning_rate": 6.2333333333333335e-06, "loss": 0.0012, "step": 7850 }, { "epoch": 38.0, "learning_rate": 6.22051282051282e-06, "loss": 0.001, "step": 7875 }, { "epoch": 38.0, "learning_rate": 6.207692307692308e-06, "loss": 0.0022, "step": 7900 }, { "epoch": 38.0, "learning_rate": 6.194871794871795e-06, "loss": 0.0024, "step": 7925 }, { "epoch": 38.01, "learning_rate": 6.1820512820512825e-06, "loss": 0.0016, "step": 7950 }, { "epoch": 38.01, "learning_rate": 6.169230769230769e-06, "loss": 0.002, "step": 7975 }, { "epoch": 38.01, "learning_rate": 6.156410256410257e-06, "loss": 0.0027, "step": 8000 }, { "epoch": 38.01, "eval_loss": 0.3210403025150299, "eval_runtime": 2278.7352, "eval_samples_per_second": 2.892, "eval_steps_per_second": 0.181, "eval_wer": 12.96123779820973, "step": 8000 }, { "epoch": 38.01, "learning_rate": 6.143589743589744e-06, "loss": 0.0012, "step": 8025 }, { "epoch": 39.0, "learning_rate": 6.1307692307692315e-06, "loss": 0.0017, "step": 8050 }, { "epoch": 39.0, "learning_rate": 6.117948717948718e-06, "loss": 0.0019, "step": 8075 }, { "epoch": 39.0, "learning_rate": 6.105128205128206e-06, "loss": 0.0035, "step": 8100 }, { "epoch": 39.0, "learning_rate": 6.092307692307693e-06, "loss": 0.0044, "step": 8125 }, { "epoch": 39.01, "learning_rate": 6.0794871794871805e-06, "loss": 0.0024, "step": 8150 }, { "epoch": 39.01, "learning_rate": 6.066666666666667e-06, "loss": 0.0023, "step": 8175 }, { "epoch": 39.01, "learning_rate": 6.053846153846155e-06, "loss": 0.0024, "step": 8200 }, { "epoch": 39.01, "learning_rate": 6.041025641025642e-06, "loss": 0.0024, "step": 8225 }, { "epoch": 40.0, "learning_rate": 6.028205128205129e-06, "loss": 0.0031, "step": 8250 }, { "epoch": 40.0, "learning_rate": 6.015384615384616e-06, "loss": 0.0026, "step": 8275 }, { "epoch": 40.0, "learning_rate": 6.002564102564103e-06, "loss": 0.003, "step": 8300 }, { "epoch": 40.0, "learning_rate": 5.989743589743591e-06, "loss": 0.0026, "step": 8325 }, { "epoch": 40.01, "learning_rate": 5.976923076923078e-06, "loss": 0.0017, "step": 8350 }, { "epoch": 40.01, "learning_rate": 5.9641025641025644e-06, "loss": 0.0015, "step": 8375 }, { "epoch": 40.01, "learning_rate": 5.951282051282051e-06, "loss": 0.0013, "step": 8400 }, { "epoch": 40.01, "learning_rate": 5.9389743589743595e-06, "loss": 0.0027, "step": 8425 }, { "epoch": 41.0, "learning_rate": 5.926153846153847e-06, "loss": 0.0024, "step": 8450 }, { "epoch": 41.0, "learning_rate": 5.913333333333334e-06, "loss": 0.0021, "step": 8475 }, { "epoch": 41.0, "learning_rate": 5.900512820512822e-06, "loss": 0.0016, "step": 8500 }, { "epoch": 41.0, "learning_rate": 5.8876923076923085e-06, "loss": 0.0017, "step": 8525 }, { "epoch": 41.01, "learning_rate": 5.8748717948717945e-06, "loss": 0.0015, "step": 8550 }, { "epoch": 41.01, "learning_rate": 5.862051282051282e-06, "loss": 0.0014, "step": 8575 }, { "epoch": 41.01, "learning_rate": 5.849230769230769e-06, "loss": 0.0012, "step": 8600 }, { "epoch": 41.01, "learning_rate": 5.836410256410257e-06, "loss": 0.0017, "step": 8625 }, { "epoch": 41.01, "learning_rate": 5.8235897435897435e-06, "loss": 0.0015, "step": 8650 }, { "epoch": 42.0, "learning_rate": 5.810769230769231e-06, "loss": 0.0014, "step": 8675 }, { "epoch": 42.0, "learning_rate": 5.797948717948718e-06, "loss": 0.0013, "step": 8700 }, { "epoch": 42.0, "learning_rate": 5.785128205128206e-06, "loss": 0.0008, "step": 8725 }, { "epoch": 42.0, "learning_rate": 5.7723076923076925e-06, "loss": 0.001, "step": 8750 }, { "epoch": 42.01, "learning_rate": 5.75948717948718e-06, "loss": 0.0016, "step": 8775 }, { "epoch": 42.01, "learning_rate": 5.746666666666667e-06, "loss": 0.0013, "step": 8800 }, { "epoch": 42.01, "learning_rate": 5.733846153846155e-06, "loss": 0.0015, "step": 8825 }, { "epoch": 42.01, "learning_rate": 5.7210256410256415e-06, "loss": 0.0011, "step": 8850 }, { "epoch": 43.0, "learning_rate": 5.708205128205129e-06, "loss": 0.0012, "step": 8875 }, { "epoch": 43.0, "learning_rate": 5.695384615384616e-06, "loss": 0.0008, "step": 8900 }, { "epoch": 43.0, "learning_rate": 5.682564102564103e-06, "loss": 0.0005, "step": 8925 }, { "epoch": 43.0, "learning_rate": 5.6697435897435905e-06, "loss": 0.0003, "step": 8950 }, { "epoch": 43.01, "learning_rate": 5.656923076923077e-06, "loss": 0.0003, "step": 8975 }, { "epoch": 43.01, "learning_rate": 5.644102564102565e-06, "loss": 0.0005, "step": 9000 }, { "epoch": 43.01, "eval_loss": 0.3375512957572937, "eval_runtime": 2304.5846, "eval_samples_per_second": 2.86, "eval_steps_per_second": 0.179, "eval_wer": 12.785045971890316, "step": 9000 }, { "epoch": 43.01, "learning_rate": 5.631282051282052e-06, "loss": 0.0006, "step": 9025 }, { "epoch": 43.01, "learning_rate": 5.6184615384615395e-06, "loss": 0.0004, "step": 9050 }, { "epoch": 44.0, "learning_rate": 5.605641025641026e-06, "loss": 0.0004, "step": 9075 }, { "epoch": 44.0, "learning_rate": 5.592820512820514e-06, "loss": 0.0004, "step": 9100 }, { "epoch": 44.0, "learning_rate": 5.580000000000001e-06, "loss": 0.0006, "step": 9125 }, { "epoch": 44.0, "learning_rate": 5.567179487179487e-06, "loss": 0.0003, "step": 9150 }, { "epoch": 44.01, "learning_rate": 5.5543589743589744e-06, "loss": 0.0004, "step": 9175 }, { "epoch": 44.01, "learning_rate": 5.541538461538461e-06, "loss": 0.0003, "step": 9200 }, { "epoch": 44.01, "learning_rate": 5.528717948717949e-06, "loss": 0.0005, "step": 9225 }, { "epoch": 44.01, "learning_rate": 5.515897435897436e-06, "loss": 0.0004, "step": 9250 }, { "epoch": 45.0, "learning_rate": 5.5030769230769234e-06, "loss": 0.0005, "step": 9275 }, { "epoch": 45.0, "learning_rate": 5.49025641025641e-06, "loss": 0.0003, "step": 9300 }, { "epoch": 45.0, "learning_rate": 5.477435897435898e-06, "loss": 0.0003, "step": 9325 }, { "epoch": 45.0, "learning_rate": 5.464615384615385e-06, "loss": 0.0003, "step": 9350 }, { "epoch": 45.01, "learning_rate": 5.451794871794872e-06, "loss": 0.0004, "step": 9375 }, { "epoch": 45.01, "learning_rate": 5.438974358974359e-06, "loss": 0.0005, "step": 9400 }, { "epoch": 45.01, "learning_rate": 5.426153846153847e-06, "loss": 0.0006, "step": 9425 }, { "epoch": 45.01, "learning_rate": 5.413333333333334e-06, "loss": 0.0003, "step": 9450 }, { "epoch": 45.01, "learning_rate": 5.400512820512821e-06, "loss": 0.0005, "step": 9475 }, { "epoch": 46.0, "learning_rate": 5.387692307692308e-06, "loss": 0.0008, "step": 9500 }, { "epoch": 46.0, "learning_rate": 5.374871794871795e-06, "loss": 0.0008, "step": 9525 }, { "epoch": 46.0, "learning_rate": 5.362051282051283e-06, "loss": 0.0004, "step": 9550 }, { "epoch": 46.0, "learning_rate": 5.3492307692307696e-06, "loss": 0.0008, "step": 9575 }, { "epoch": 46.01, "learning_rate": 5.336410256410257e-06, "loss": 0.0005, "step": 9600 }, { "epoch": 46.01, "learning_rate": 5.323589743589744e-06, "loss": 0.0014, "step": 9625 }, { "epoch": 46.01, "learning_rate": 5.310769230769232e-06, "loss": 0.0206, "step": 9650 }, { "epoch": 46.01, "learning_rate": 5.2979487179487185e-06, "loss": 0.0011, "step": 9675 }, { "epoch": 47.0, "learning_rate": 5.285128205128206e-06, "loss": 0.0023, "step": 9700 }, { "epoch": 47.0, "learning_rate": 5.272307692307693e-06, "loss": 0.0021, "step": 9725 }, { "epoch": 47.0, "learning_rate": 5.259487179487181e-06, "loss": 0.0021, "step": 9750 }, { "epoch": 47.0, "learning_rate": 5.246666666666667e-06, "loss": 0.0028, "step": 9775 }, { "epoch": 47.01, "learning_rate": 5.2338461538461535e-06, "loss": 0.0027, "step": 9800 }, { "epoch": 47.01, "learning_rate": 5.221025641025641e-06, "loss": 0.0017, "step": 9825 }, { "epoch": 47.01, "learning_rate": 5.208205128205128e-06, "loss": 0.0017, "step": 9850 }, { "epoch": 47.01, "learning_rate": 5.195384615384616e-06, "loss": 0.0011, "step": 9875 }, { "epoch": 48.0, "learning_rate": 5.1825641025641025e-06, "loss": 0.0018, "step": 9900 }, { "epoch": 48.0, "learning_rate": 5.16974358974359e-06, "loss": 0.0009, "step": 9925 }, { "epoch": 48.0, "learning_rate": 5.156923076923077e-06, "loss": 0.0012, "step": 9950 }, { "epoch": 48.0, "learning_rate": 5.144102564102565e-06, "loss": 0.0012, "step": 9975 }, { "epoch": 48.01, "learning_rate": 5.1312820512820515e-06, "loss": 0.0007, "step": 10000 }, { "epoch": 48.01, "eval_loss": 0.3361070156097412, "eval_runtime": 2294.0923, "eval_samples_per_second": 2.873, "eval_steps_per_second": 0.18, "eval_wer": 12.912633156466443, "step": 10000 }, { "epoch": 48.01, "learning_rate": 5.118461538461539e-06, "loss": 0.0008, "step": 10025 }, { "epoch": 48.01, "learning_rate": 5.105641025641026e-06, "loss": 0.0007, "step": 10050 }, { "epoch": 48.01, "learning_rate": 5.092820512820514e-06, "loss": 0.0004, "step": 10075 }, { "epoch": 49.0, "learning_rate": 5.0800000000000005e-06, "loss": 0.0006, "step": 10100 }, { "epoch": 49.0, "learning_rate": 5.067179487179487e-06, "loss": 0.0004, "step": 10125 }, { "epoch": 49.0, "learning_rate": 5.054358974358975e-06, "loss": 0.0005, "step": 10150 }, { "epoch": 49.0, "learning_rate": 5.041538461538462e-06, "loss": 0.0005, "step": 10175 }, { "epoch": 49.01, "learning_rate": 5.0287179487179495e-06, "loss": 0.0002, "step": 10200 }, { "epoch": 49.01, "learning_rate": 5.015897435897436e-06, "loss": 0.0005, "step": 10225 }, { "epoch": 49.01, "learning_rate": 5.003076923076924e-06, "loss": 0.0007, "step": 10250 }, { "epoch": 49.01, "learning_rate": 4.990256410256411e-06, "loss": 0.0003, "step": 10275 }, { "epoch": 49.01, "learning_rate": 4.977435897435898e-06, "loss": 0.0004, "step": 10300 }, { "epoch": 50.0, "learning_rate": 4.964615384615385e-06, "loss": 0.0005, "step": 10325 }, { "epoch": 50.0, "learning_rate": 4.951794871794872e-06, "loss": 0.0006, "step": 10350 }, { "epoch": 50.0, "learning_rate": 4.93897435897436e-06, "loss": 0.0004, "step": 10375 }, { "epoch": 50.01, "learning_rate": 4.926153846153847e-06, "loss": 0.0007, "step": 10400 }, { "epoch": 50.01, "learning_rate": 4.9133333333333334e-06, "loss": 0.0012, "step": 10425 }, { "epoch": 50.01, "learning_rate": 4.900512820512821e-06, "loss": 0.0006, "step": 10450 }, { "epoch": 50.01, "learning_rate": 4.887692307692308e-06, "loss": 0.0007, "step": 10475 }, { "epoch": 50.01, "learning_rate": 4.874871794871796e-06, "loss": 0.0008, "step": 10500 }, { "epoch": 51.0, "learning_rate": 4.8620512820512824e-06, "loss": 0.0008, "step": 10525 }, { "epoch": 51.0, "learning_rate": 4.849230769230769e-06, "loss": 0.0014, "step": 10550 }, { "epoch": 51.0, "learning_rate": 4.836410256410257e-06, "loss": 0.0006, "step": 10575 }, { "epoch": 51.0, "learning_rate": 4.823589743589744e-06, "loss": 0.0005, "step": 10600 }, { "epoch": 51.01, "learning_rate": 4.8107692307692314e-06, "loss": 0.0004, "step": 10625 }, { "epoch": 51.01, "learning_rate": 4.797948717948718e-06, "loss": 0.0002, "step": 10650 }, { "epoch": 51.01, "learning_rate": 4.785128205128206e-06, "loss": 0.0002, "step": 10675 }, { "epoch": 51.01, "learning_rate": 4.772307692307693e-06, "loss": 0.001, "step": 10700 }, { "epoch": 52.0, "learning_rate": 4.7594871794871796e-06, "loss": 0.0005, "step": 10725 }, { "epoch": 52.0, "learning_rate": 4.746666666666667e-06, "loss": 0.0006, "step": 10750 }, { "epoch": 52.0, "learning_rate": 4.733846153846154e-06, "loss": 0.0006, "step": 10775 }, { "epoch": 52.0, "learning_rate": 4.721025641025642e-06, "loss": 0.0002, "step": 10800 }, { "epoch": 52.01, "learning_rate": 4.7082051282051286e-06, "loss": 0.0002, "step": 10825 }, { "epoch": 52.01, "learning_rate": 4.695384615384615e-06, "loss": 0.0002, "step": 10850 }, { "epoch": 52.01, "learning_rate": 4.682564102564103e-06, "loss": 0.0005, "step": 10875 }, { "epoch": 52.01, "learning_rate": 4.66974358974359e-06, "loss": 0.0003, "step": 10900 }, { "epoch": 53.0, "learning_rate": 4.6569230769230776e-06, "loss": 0.0005, "step": 10925 }, { "epoch": 53.0, "learning_rate": 4.644102564102564e-06, "loss": 0.0004, "step": 10950 }, { "epoch": 53.0, "learning_rate": 4.631282051282052e-06, "loss": 0.0003, "step": 10975 }, { "epoch": 53.0, "learning_rate": 4.618461538461539e-06, "loss": 0.0002, "step": 11000 }, { "epoch": 53.0, "eval_loss": 0.3559357225894928, "eval_runtime": 2313.8828, "eval_samples_per_second": 2.848, "eval_steps_per_second": 0.178, "eval_wer": 12.37393171047835, "step": 11000 }, { "epoch": 53.01, "learning_rate": 4.605641025641026e-06, "loss": 0.0001, "step": 11025 }, { "epoch": 53.01, "learning_rate": 4.592820512820513e-06, "loss": 0.0006, "step": 11050 }, { "epoch": 53.01, "learning_rate": 4.58e-06, "loss": 0.0004, "step": 11075 }, { "epoch": 53.01, "learning_rate": 4.567179487179488e-06, "loss": 0.0003, "step": 11100 }, { "epoch": 54.0, "learning_rate": 4.554358974358975e-06, "loss": 0.0011, "step": 11125 }, { "epoch": 54.0, "learning_rate": 4.5415384615384615e-06, "loss": 0.0005, "step": 11150 }, { "epoch": 54.0, "learning_rate": 4.528717948717949e-06, "loss": 0.0003, "step": 11175 }, { "epoch": 54.0, "learning_rate": 4.515897435897436e-06, "loss": 0.0002, "step": 11200 }, { "epoch": 54.01, "learning_rate": 4.503076923076924e-06, "loss": 0.0002, "step": 11225 }, { "epoch": 54.01, "learning_rate": 4.4902564102564105e-06, "loss": 0.0004, "step": 11250 }, { "epoch": 54.01, "learning_rate": 4.477435897435898e-06, "loss": 0.0007, "step": 11275 }, { "epoch": 54.01, "learning_rate": 4.464615384615385e-06, "loss": 0.0012, "step": 11300 }, { "epoch": 54.01, "learning_rate": 4.451794871794872e-06, "loss": 0.001, "step": 11325 }, { "epoch": 55.0, "learning_rate": 4.4389743589743595e-06, "loss": 0.0013, "step": 11350 }, { "epoch": 55.0, "learning_rate": 4.426153846153846e-06, "loss": 0.0005, "step": 11375 }, { "epoch": 55.0, "learning_rate": 4.413333333333334e-06, "loss": 0.001, "step": 11400 }, { "epoch": 55.0, "learning_rate": 4.400512820512821e-06, "loss": 0.001, "step": 11425 }, { "epoch": 55.01, "learning_rate": 4.387692307692308e-06, "loss": 0.0018, "step": 11450 }, { "epoch": 55.01, "learning_rate": 4.374871794871795e-06, "loss": 0.0009, "step": 11475 }, { "epoch": 55.01, "learning_rate": 4.362051282051282e-06, "loss": 0.0009, "step": 11500 }, { "epoch": 55.01, "learning_rate": 4.34923076923077e-06, "loss": 0.0017, "step": 11525 }, { "epoch": 56.0, "learning_rate": 4.336410256410257e-06, "loss": 0.0019, "step": 11550 }, { "epoch": 56.0, "learning_rate": 4.323589743589744e-06, "loss": 0.0016, "step": 11575 }, { "epoch": 56.0, "learning_rate": 4.310769230769231e-06, "loss": 0.0012, "step": 11600 }, { "epoch": 56.0, "learning_rate": 4.297948717948718e-06, "loss": 0.0006, "step": 11625 }, { "epoch": 56.01, "learning_rate": 4.285128205128206e-06, "loss": 0.001, "step": 11650 }, { "epoch": 56.01, "learning_rate": 4.2723076923076925e-06, "loss": 0.0011, "step": 11675 }, { "epoch": 56.01, "learning_rate": 4.25948717948718e-06, "loss": 0.0008, "step": 11700 }, { "epoch": 56.01, "learning_rate": 4.246666666666667e-06, "loss": 0.0005, "step": 11725 }, { "epoch": 57.0, "learning_rate": 4.233846153846154e-06, "loss": 0.0002, "step": 11750 }, { "epoch": 57.0, "learning_rate": 4.2210256410256414e-06, "loss": 0.0003, "step": 11775 }, { "epoch": 57.0, "learning_rate": 4.208205128205128e-06, "loss": 0.0006, "step": 11800 }, { "epoch": 57.0, "learning_rate": 4.195384615384616e-06, "loss": 0.0006, "step": 11825 }, { "epoch": 57.01, "learning_rate": 4.182564102564103e-06, "loss": 0.0002, "step": 11850 }, { "epoch": 57.01, "learning_rate": 4.1697435897435904e-06, "loss": 0.0002, "step": 11875 }, { "epoch": 57.01, "learning_rate": 4.156923076923077e-06, "loss": 0.0001, "step": 11900 }, { "epoch": 57.01, "learning_rate": 4.144102564102564e-06, "loss": 0.0001, "step": 11925 }, { "epoch": 58.0, "learning_rate": 4.131282051282052e-06, "loss": 0.0001, "step": 11950 }, { "epoch": 58.0, "learning_rate": 4.118461538461539e-06, "loss": 0.0001, "step": 11975 }, { "epoch": 58.0, "learning_rate": 4.105641025641026e-06, "loss": 0.0001, "step": 12000 }, { "epoch": 58.0, "eval_loss": 0.3550183176994324, "eval_runtime": 2312.7367, "eval_samples_per_second": 2.85, "eval_steps_per_second": 0.178, "eval_wer": 12.335453035764916, "step": 12000 }, { "epoch": 58.0, "learning_rate": 4.092820512820513e-06, "loss": 0.0004, "step": 12025 }, { "epoch": 58.01, "learning_rate": 4.08e-06, "loss": 0.0003, "step": 12050 }, { "epoch": 58.01, "learning_rate": 4.0671794871794876e-06, "loss": 0.0002, "step": 12075 }, { "epoch": 58.01, "learning_rate": 4.054358974358974e-06, "loss": 0.0001, "step": 12100 }, { "epoch": 58.01, "learning_rate": 4.041538461538462e-06, "loss": 0.0004, "step": 12125 }, { "epoch": 58.01, "learning_rate": 4.028717948717949e-06, "loss": 0.0001, "step": 12150 }, { "epoch": 59.0, "learning_rate": 4.0158974358974366e-06, "loss": 0.0001, "step": 12175 }, { "epoch": 59.0, "learning_rate": 4.003076923076923e-06, "loss": 0.0001, "step": 12200 }, { "epoch": 59.0, "learning_rate": 3.990256410256411e-06, "loss": 0.0001, "step": 12225 }, { "epoch": 59.0, "learning_rate": 3.977435897435898e-06, "loss": 0.0001, "step": 12250 }, { "epoch": 59.01, "learning_rate": 3.964615384615385e-06, "loss": 0.0001, "step": 12275 }, { "epoch": 59.01, "learning_rate": 3.951794871794872e-06, "loss": 0.0001, "step": 12300 }, { "epoch": 59.01, "learning_rate": 3.938974358974359e-06, "loss": 0.0001, "step": 12325 }, { "epoch": 59.01, "learning_rate": 3.926153846153846e-06, "loss": 0.0001, "step": 12350 }, { "epoch": 60.0, "learning_rate": 3.913333333333334e-06, "loss": 0.0001, "step": 12375 }, { "epoch": 60.0, "learning_rate": 3.9005128205128205e-06, "loss": 0.0001, "step": 12400 }, { "epoch": 60.0, "learning_rate": 3.887692307692308e-06, "loss": 0.0001, "step": 12425 }, { "epoch": 60.0, "learning_rate": 3.874871794871795e-06, "loss": 0.0001, "step": 12450 }, { "epoch": 60.01, "learning_rate": 3.862051282051283e-06, "loss": 0.0001, "step": 12475 }, { "epoch": 60.01, "learning_rate": 3.8492307692307695e-06, "loss": 0.0001, "step": 12500 }, { "epoch": 60.01, "learning_rate": 3.836410256410257e-06, "loss": 0.0001, "step": 12525 }, { "epoch": 60.01, "learning_rate": 3.823589743589744e-06, "loss": 0.0001, "step": 12550 }, { "epoch": 61.0, "learning_rate": 3.8107692307692313e-06, "loss": 0.0001, "step": 12575 }, { "epoch": 61.0, "learning_rate": 3.7979487179487185e-06, "loss": 0.0001, "step": 12600 }, { "epoch": 61.0, "learning_rate": 3.7851282051282058e-06, "loss": 0.0001, "step": 12625 }, { "epoch": 61.0, "learning_rate": 3.772307692307693e-06, "loss": 0.0001, "step": 12650 }, { "epoch": 61.01, "learning_rate": 3.7594871794871794e-06, "loss": 0.0001, "step": 12675 }, { "epoch": 61.01, "learning_rate": 3.7466666666666667e-06, "loss": 0.0001, "step": 12700 }, { "epoch": 61.01, "learning_rate": 3.733846153846154e-06, "loss": 0.0001, "step": 12725 }, { "epoch": 61.01, "learning_rate": 3.721025641025641e-06, "loss": 0.0001, "step": 12750 }, { "epoch": 62.0, "learning_rate": 3.7082051282051284e-06, "loss": 0.0001, "step": 12775 }, { "epoch": 62.0, "learning_rate": 3.6953846153846156e-06, "loss": 0.0001, "step": 12800 }, { "epoch": 62.0, "learning_rate": 3.682564102564103e-06, "loss": 0.0001, "step": 12825 }, { "epoch": 62.0, "learning_rate": 3.66974358974359e-06, "loss": 0.0001, "step": 12850 }, { "epoch": 62.01, "learning_rate": 3.6569230769230774e-06, "loss": 0.0001, "step": 12875 }, { "epoch": 62.01, "learning_rate": 3.6441025641025646e-06, "loss": 0.0001, "step": 12900 }, { "epoch": 62.01, "learning_rate": 3.631282051282052e-06, "loss": 0.0, "step": 12925 }, { "epoch": 62.01, "learning_rate": 3.618461538461539e-06, "loss": 0.0, "step": 12950 }, { "epoch": 62.01, "learning_rate": 3.6056410256410255e-06, "loss": 0.0, "step": 12975 }, { "epoch": 63.0, "learning_rate": 3.5928205128205128e-06, "loss": 0.0, "step": 13000 }, { "epoch": 63.0, "eval_loss": 0.3851548433303833, "eval_runtime": 2269.3953, "eval_samples_per_second": 2.904, "eval_steps_per_second": 0.182, "eval_wer": 12.114706954514157, "step": 13000 }, { "epoch": 63.0, "learning_rate": 3.58e-06, "loss": 0.0, "step": 13025 }, { "epoch": 63.0, "learning_rate": 3.5671794871794873e-06, "loss": 0.0, "step": 13050 }, { "epoch": 63.0, "learning_rate": 3.5543589743589745e-06, "loss": 0.0, "step": 13075 }, { "epoch": 63.01, "learning_rate": 3.5415384615384618e-06, "loss": 0.0, "step": 13100 }, { "epoch": 63.01, "learning_rate": 3.528717948717949e-06, "loss": 0.0, "step": 13125 }, { "epoch": 63.01, "learning_rate": 3.5158974358974363e-06, "loss": 0.0, "step": 13150 }, { "epoch": 63.01, "learning_rate": 3.5030769230769235e-06, "loss": 0.0, "step": 13175 }, { "epoch": 64.0, "learning_rate": 3.4902564102564108e-06, "loss": 0.0, "step": 13200 }, { "epoch": 64.0, "learning_rate": 3.477435897435898e-06, "loss": 0.0, "step": 13225 }, { "epoch": 64.0, "learning_rate": 3.4646153846153853e-06, "loss": 0.0, "step": 13250 }, { "epoch": 64.0, "learning_rate": 3.4517948717948717e-06, "loss": 0.0, "step": 13275 }, { "epoch": 64.01, "learning_rate": 3.438974358974359e-06, "loss": 0.0, "step": 13300 }, { "epoch": 64.01, "learning_rate": 3.426153846153846e-06, "loss": 0.0, "step": 13325 }, { "epoch": 64.01, "learning_rate": 3.4133333333333334e-06, "loss": 0.0, "step": 13350 }, { "epoch": 64.01, "learning_rate": 3.4005128205128207e-06, "loss": 0.0, "step": 13375 }, { "epoch": 65.0, "learning_rate": 3.387692307692308e-06, "loss": 0.0, "step": 13400 }, { "epoch": 65.0, "learning_rate": 3.374871794871795e-06, "loss": 0.0, "step": 13425 }, { "epoch": 65.0, "learning_rate": 3.3620512820512824e-06, "loss": 0.0, "step": 13450 }, { "epoch": 65.0, "learning_rate": 3.3492307692307696e-06, "loss": 0.0, "step": 13475 }, { "epoch": 65.01, "learning_rate": 3.336410256410257e-06, "loss": 0.0, "step": 13500 }, { "epoch": 65.01, "learning_rate": 3.323589743589744e-06, "loss": 0.0, "step": 13525 }, { "epoch": 65.01, "learning_rate": 3.3107692307692314e-06, "loss": 0.0, "step": 13550 }, { "epoch": 65.01, "learning_rate": 3.297948717948718e-06, "loss": 0.0, "step": 13575 }, { "epoch": 66.0, "learning_rate": 3.285128205128205e-06, "loss": 0.0, "step": 13600 }, { "epoch": 66.0, "learning_rate": 3.2723076923076923e-06, "loss": 0.0, "step": 13625 }, { "epoch": 66.0, "learning_rate": 3.2594871794871795e-06, "loss": 0.0, "step": 13650 }, { "epoch": 66.0, "learning_rate": 3.2466666666666668e-06, "loss": 0.0, "step": 13675 }, { "epoch": 66.01, "learning_rate": 3.233846153846154e-06, "loss": 0.0, "step": 13700 }, { "epoch": 66.01, "learning_rate": 3.2210256410256413e-06, "loss": 0.0, "step": 13725 }, { "epoch": 66.01, "learning_rate": 3.2082051282051285e-06, "loss": 0.0, "step": 13750 }, { "epoch": 66.01, "learning_rate": 3.1953846153846158e-06, "loss": 0.0, "step": 13775 }, { "epoch": 66.01, "learning_rate": 3.182564102564103e-06, "loss": 0.0, "step": 13800 }, { "epoch": 67.0, "learning_rate": 3.1697435897435903e-06, "loss": 0.0, "step": 13825 }, { "epoch": 67.0, "learning_rate": 3.1569230769230775e-06, "loss": 0.0, "step": 13850 }, { "epoch": 67.0, "learning_rate": 3.144102564102564e-06, "loss": 0.0, "step": 13875 }, { "epoch": 67.0, "learning_rate": 3.131282051282051e-06, "loss": 0.0, "step": 13900 }, { "epoch": 67.01, "learning_rate": 3.1184615384615384e-06, "loss": 0.0, "step": 13925 }, { "epoch": 67.01, "learning_rate": 3.1056410256410257e-06, "loss": 0.0, "step": 13950 }, { "epoch": 67.01, "learning_rate": 3.092820512820513e-06, "loss": 0.0, "step": 13975 }, { "epoch": 67.01, "learning_rate": 3.08e-06, "loss": 0.0, "step": 14000 }, { "epoch": 67.01, "eval_loss": 0.3973710834980011, "eval_runtime": 2273.8662, "eval_samples_per_second": 2.899, "eval_steps_per_second": 0.181, "eval_wer": 12.013447284215642, "step": 14000 }, { "epoch": 68.0, "learning_rate": 3.0671794871794874e-06, "loss": 0.0, "step": 14025 }, { "epoch": 68.0, "learning_rate": 3.0543589743589747e-06, "loss": 0.0, "step": 14050 }, { "epoch": 68.0, "learning_rate": 3.041538461538462e-06, "loss": 0.0, "step": 14075 }, { "epoch": 68.0, "learning_rate": 3.028717948717949e-06, "loss": 0.0, "step": 14100 }, { "epoch": 68.01, "learning_rate": 3.0158974358974364e-06, "loss": 0.0, "step": 14125 }, { "epoch": 68.01, "learning_rate": 3.0030769230769236e-06, "loss": 0.0, "step": 14150 }, { "epoch": 68.01, "learning_rate": 2.99025641025641e-06, "loss": 0.0, "step": 14175 }, { "epoch": 68.01, "learning_rate": 2.9774358974358973e-06, "loss": 0.0, "step": 14200 }, { "epoch": 69.0, "learning_rate": 2.9646153846153845e-06, "loss": 0.0, "step": 14225 }, { "epoch": 69.0, "learning_rate": 2.951794871794872e-06, "loss": 0.0, "step": 14250 }, { "epoch": 69.0, "learning_rate": 2.938974358974359e-06, "loss": 0.0, "step": 14275 }, { "epoch": 69.0, "learning_rate": 2.9261538461538463e-06, "loss": 0.0, "step": 14300 }, { "epoch": 69.01, "learning_rate": 2.9133333333333335e-06, "loss": 0.0, "step": 14325 }, { "epoch": 69.01, "learning_rate": 2.9005128205128208e-06, "loss": 0.0, "step": 14350 }, { "epoch": 69.01, "learning_rate": 2.887692307692308e-06, "loss": 0.0, "step": 14375 }, { "epoch": 69.01, "learning_rate": 2.8748717948717953e-06, "loss": 0.0, "step": 14400 }, { "epoch": 70.0, "learning_rate": 2.8620512820512825e-06, "loss": 0.0, "step": 14425 }, { "epoch": 70.0, "learning_rate": 2.8492307692307698e-06, "loss": 0.0, "step": 14450 }, { "epoch": 70.0, "learning_rate": 2.836410256410257e-06, "loss": 0.0, "step": 14475 }, { "epoch": 70.0, "learning_rate": 2.8235897435897434e-06, "loss": 0.0, "step": 14500 }, { "epoch": 70.01, "learning_rate": 2.8107692307692307e-06, "loss": 0.0, "step": 14525 }, { "epoch": 70.01, "learning_rate": 2.797948717948718e-06, "loss": 0.0, "step": 14550 }, { "epoch": 70.01, "learning_rate": 2.785128205128205e-06, "loss": 0.0, "step": 14575 }, { "epoch": 70.01, "learning_rate": 2.7723076923076924e-06, "loss": 0.0, "step": 14600 }, { "epoch": 70.01, "learning_rate": 2.7594871794871797e-06, "loss": 0.0, "step": 14625 }, { "epoch": 71.0, "learning_rate": 2.746666666666667e-06, "loss": 0.0, "step": 14650 }, { "epoch": 71.0, "learning_rate": 2.733846153846154e-06, "loss": 0.0, "step": 14675 }, { "epoch": 71.0, "learning_rate": 2.7210256410256414e-06, "loss": 0.0, "step": 14700 }, { "epoch": 71.0, "learning_rate": 2.7082051282051287e-06, "loss": 0.0, "step": 14725 }, { "epoch": 71.01, "learning_rate": 2.695384615384616e-06, "loss": 0.0, "step": 14750 }, { "epoch": 71.01, "learning_rate": 2.682564102564103e-06, "loss": 0.0, "step": 14775 }, { "epoch": 71.01, "learning_rate": 2.6697435897435896e-06, "loss": 0.0, "step": 14800 }, { "epoch": 71.01, "learning_rate": 2.656923076923077e-06, "loss": 0.0, "step": 14825 }, { "epoch": 72.0, "learning_rate": 2.644102564102564e-06, "loss": 0.0, "step": 14850 }, { "epoch": 72.0, "learning_rate": 2.6312820512820513e-06, "loss": 0.0, "step": 14875 }, { "epoch": 72.0, "learning_rate": 2.6184615384615385e-06, "loss": 0.0, "step": 14900 }, { "epoch": 72.0, "learning_rate": 2.605641025641026e-06, "loss": 0.0, "step": 14925 }, { "epoch": 72.01, "learning_rate": 2.592820512820513e-06, "loss": 0.0, "step": 14950 }, { "epoch": 72.01, "learning_rate": 2.5800000000000003e-06, "loss": 0.0, "step": 14975 }, { "epoch": 72.01, "learning_rate": 2.5671794871794875e-06, "loss": 0.0, "step": 15000 }, { "epoch": 72.01, "eval_loss": 0.407157301902771, "eval_runtime": 2262.2649, "eval_samples_per_second": 2.913, "eval_steps_per_second": 0.182, "eval_wer": 11.944590708412653, "step": 15000 }, { "epoch": 72.01, "learning_rate": 2.5543589743589748e-06, "loss": 0.0, "step": 15025 }, { "epoch": 73.0, "learning_rate": 2.541538461538462e-06, "loss": 0.0, "step": 15050 }, { "epoch": 73.0, "learning_rate": 2.5287179487179493e-06, "loss": 0.0, "step": 15075 }, { "epoch": 73.0, "learning_rate": 2.5158974358974357e-06, "loss": 0.0, "step": 15100 }, { "epoch": 73.0, "learning_rate": 2.503076923076923e-06, "loss": 0.0, "step": 15125 }, { "epoch": 73.01, "learning_rate": 2.4902564102564106e-06, "loss": 0.0, "step": 15150 }, { "epoch": 73.01, "learning_rate": 2.4774358974358974e-06, "loss": 0.0, "step": 15175 }, { "epoch": 73.01, "learning_rate": 2.4646153846153847e-06, "loss": 0.0, "step": 15200 }, { "epoch": 73.01, "learning_rate": 2.451794871794872e-06, "loss": 0.0, "step": 15225 }, { "epoch": 74.0, "learning_rate": 2.438974358974359e-06, "loss": 0.0, "step": 15250 }, { "epoch": 74.0, "learning_rate": 2.4261538461538464e-06, "loss": 0.0, "step": 15275 }, { "epoch": 74.0, "learning_rate": 2.4133333333333337e-06, "loss": 0.0, "step": 15300 }, { "epoch": 74.0, "learning_rate": 2.4005128205128205e-06, "loss": 0.0, "step": 15325 }, { "epoch": 74.01, "learning_rate": 2.3876923076923077e-06, "loss": 0.0, "step": 15350 }, { "epoch": 74.01, "learning_rate": 2.374871794871795e-06, "loss": 0.0, "step": 15375 }, { "epoch": 74.01, "learning_rate": 2.3620512820512822e-06, "loss": 0.0, "step": 15400 }, { "epoch": 74.01, "learning_rate": 2.3492307692307695e-06, "loss": 0.0, "step": 15425 }, { "epoch": 74.01, "learning_rate": 2.3364102564102567e-06, "loss": 0.0, "step": 15450 }, { "epoch": 75.0, "learning_rate": 2.3235897435897436e-06, "loss": 0.0, "step": 15475 }, { "epoch": 75.0, "learning_rate": 2.310769230769231e-06, "loss": 0.0, "step": 15500 }, { "epoch": 75.0, "learning_rate": 2.297948717948718e-06, "loss": 0.0, "step": 15525 }, { "epoch": 75.0, "learning_rate": 2.2851282051282053e-06, "loss": 0.0, "step": 15550 }, { "epoch": 75.01, "learning_rate": 2.2723076923076925e-06, "loss": 0.0, "step": 15575 }, { "epoch": 75.01, "learning_rate": 2.25948717948718e-06, "loss": 0.0, "step": 15600 }, { "epoch": 75.01, "learning_rate": 2.2466666666666666e-06, "loss": 0.0, "step": 15625 }, { "epoch": 75.01, "learning_rate": 2.233846153846154e-06, "loss": 0.0, "step": 15650 }, { "epoch": 76.0, "learning_rate": 2.221025641025641e-06, "loss": 0.0, "step": 15675 }, { "epoch": 76.0, "learning_rate": 2.2082051282051284e-06, "loss": 0.0, "step": 15700 }, { "epoch": 76.0, "learning_rate": 2.1953846153846156e-06, "loss": 0.0, "step": 15725 }, { "epoch": 76.0, "learning_rate": 2.182564102564103e-06, "loss": 0.0, "step": 15750 }, { "epoch": 76.01, "learning_rate": 2.1697435897435897e-06, "loss": 0.0, "step": 15775 }, { "epoch": 76.01, "learning_rate": 2.156923076923077e-06, "loss": 0.0, "step": 15800 }, { "epoch": 76.01, "learning_rate": 2.144102564102564e-06, "loss": 0.0, "step": 15825 }, { "epoch": 76.01, "learning_rate": 2.1312820512820514e-06, "loss": 0.0, "step": 15850 }, { "epoch": 77.0, "learning_rate": 2.1184615384615387e-06, "loss": 0.0, "step": 15875 }, { "epoch": 77.0, "learning_rate": 2.105641025641026e-06, "loss": 0.0, "step": 15900 }, { "epoch": 77.0, "learning_rate": 2.092820512820513e-06, "loss": 0.0, "step": 15925 }, { "epoch": 77.0, "learning_rate": 2.08e-06, "loss": 0.0, "step": 15950 }, { "epoch": 77.01, "learning_rate": 2.0671794871794872e-06, "loss": 0.0, "step": 15975 }, { "epoch": 77.01, "learning_rate": 2.0543589743589745e-06, "loss": 0.0, "step": 16000 }, { "epoch": 77.01, "eval_loss": 0.41622260212898254, "eval_runtime": 2273.3629, "eval_samples_per_second": 2.899, "eval_steps_per_second": 0.181, "eval_wer": 11.92028838754101, "step": 16000 }, { "epoch": 77.01, "learning_rate": 2.0415384615384617e-06, "loss": 0.0, "step": 16025 }, { "epoch": 77.01, "learning_rate": 2.028717948717949e-06, "loss": 0.0, "step": 16050 }, { "epoch": 78.0, "learning_rate": 2.0158974358974362e-06, "loss": 0.0, "step": 16075 }, { "epoch": 78.0, "learning_rate": 2.003076923076923e-06, "loss": 0.0, "step": 16100 }, { "epoch": 78.0, "learning_rate": 1.9902564102564103e-06, "loss": 0.0, "step": 16125 }, { "epoch": 78.0, "learning_rate": 1.9774358974358976e-06, "loss": 0.0, "step": 16150 }, { "epoch": 78.01, "learning_rate": 1.964615384615385e-06, "loss": 0.0, "step": 16175 }, { "epoch": 78.01, "learning_rate": 1.951794871794872e-06, "loss": 0.0, "step": 16200 }, { "epoch": 78.01, "learning_rate": 1.9389743589743593e-06, "loss": 0.0, "step": 16225 }, { "epoch": 78.01, "learning_rate": 1.926153846153846e-06, "loss": 0.0, "step": 16250 }, { "epoch": 79.0, "learning_rate": 1.9133333333333334e-06, "loss": 0.0, "step": 16275 }, { "epoch": 79.0, "learning_rate": 1.9005128205128206e-06, "loss": 0.0, "step": 16300 }, { "epoch": 79.0, "learning_rate": 1.8876923076923079e-06, "loss": 0.0, "step": 16325 }, { "epoch": 79.0, "learning_rate": 1.8748717948717951e-06, "loss": 0.0, "step": 16350 }, { "epoch": 79.01, "learning_rate": 1.8620512820512824e-06, "loss": 0.0, "step": 16375 }, { "epoch": 79.01, "learning_rate": 1.8492307692307692e-06, "loss": 0.0, "step": 16400 }, { "epoch": 79.01, "learning_rate": 1.8364102564102564e-06, "loss": 0.0, "step": 16425 }, { "epoch": 79.01, "learning_rate": 1.8235897435897437e-06, "loss": 0.0, "step": 16450 }, { "epoch": 79.01, "learning_rate": 1.810769230769231e-06, "loss": 0.0, "step": 16475 }, { "epoch": 80.0, "learning_rate": 1.7979487179487182e-06, "loss": 0.0, "step": 16500 }, { "epoch": 80.0, "learning_rate": 1.7851282051282054e-06, "loss": 0.0, "step": 16525 }, { "epoch": 80.0, "learning_rate": 1.7723076923076922e-06, "loss": 0.0, "step": 16550 }, { "epoch": 80.0, "learning_rate": 1.7594871794871795e-06, "loss": 0.0, "step": 16575 }, { "epoch": 80.01, "learning_rate": 1.7466666666666667e-06, "loss": 0.0, "step": 16600 }, { "epoch": 80.01, "learning_rate": 1.733846153846154e-06, "loss": 0.0, "step": 16625 }, { "epoch": 80.01, "learning_rate": 1.7210256410256412e-06, "loss": 0.0, "step": 16650 }, { "epoch": 80.01, "learning_rate": 1.7082051282051285e-06, "loss": 0.0, "step": 16675 }, { "epoch": 81.0, "learning_rate": 1.6953846153846153e-06, "loss": 0.0, "step": 16700 }, { "epoch": 81.0, "learning_rate": 1.6825641025641026e-06, "loss": 0.0, "step": 16725 }, { "epoch": 81.0, "learning_rate": 1.6697435897435898e-06, "loss": 0.0, "step": 16750 }, { "epoch": 81.0, "learning_rate": 1.656923076923077e-06, "loss": 0.0, "step": 16775 }, { "epoch": 81.01, "learning_rate": 1.6441025641025643e-06, "loss": 0.0, "step": 16800 }, { "epoch": 81.01, "learning_rate": 1.6312820512820516e-06, "loss": 0.0, "step": 16825 }, { "epoch": 81.01, "learning_rate": 1.6184615384615384e-06, "loss": 0.0, "step": 16850 }, { "epoch": 81.01, "learning_rate": 1.6056410256410256e-06, "loss": 0.0, "step": 16875 }, { "epoch": 82.0, "learning_rate": 1.5928205128205129e-06, "loss": 0.0, "step": 16900 }, { "epoch": 82.0, "learning_rate": 1.5800000000000001e-06, "loss": 0.0, "step": 16925 }, { "epoch": 82.0, "learning_rate": 1.5671794871794874e-06, "loss": 0.0, "step": 16950 }, { "epoch": 82.0, "learning_rate": 1.5543589743589746e-06, "loss": 0.0, "step": 16975 }, { "epoch": 82.01, "learning_rate": 1.5415384615384614e-06, "loss": 0.0, "step": 17000 }, { "epoch": 82.01, "eval_loss": 0.42448174953460693, "eval_runtime": 2316.7783, "eval_samples_per_second": 2.845, "eval_steps_per_second": 0.178, "eval_wer": 11.839280651302198, "step": 17000 }, { "epoch": 82.01, "learning_rate": 1.5287179487179487e-06, "loss": 0.0, "step": 17025 }, { "epoch": 82.01, "learning_rate": 1.515897435897436e-06, "loss": 0.0, "step": 17050 }, { "epoch": 82.01, "learning_rate": 1.5030769230769232e-06, "loss": 0.0, "step": 17075 }, { "epoch": 83.0, "learning_rate": 1.4902564102564104e-06, "loss": 0.0, "step": 17100 }, { "epoch": 83.0, "learning_rate": 1.4774358974358977e-06, "loss": 0.0, "step": 17125 }, { "epoch": 83.0, "learning_rate": 1.4646153846153845e-06, "loss": 0.0, "step": 17150 }, { "epoch": 83.0, "learning_rate": 1.4517948717948718e-06, "loss": 0.0, "step": 17175 }, { "epoch": 83.01, "learning_rate": 1.438974358974359e-06, "loss": 0.0, "step": 17200 }, { "epoch": 83.01, "learning_rate": 1.4261538461538462e-06, "loss": 0.0, "step": 17225 }, { "epoch": 83.01, "learning_rate": 1.4133333333333335e-06, "loss": 0.0, "step": 17250 }, { "epoch": 83.01, "learning_rate": 1.4005128205128207e-06, "loss": 0.0, "step": 17275 }, { "epoch": 83.01, "learning_rate": 1.3876923076923076e-06, "loss": 0.0, "step": 17300 }, { "epoch": 84.0, "learning_rate": 1.3748717948717948e-06, "loss": 0.0, "step": 17325 }, { "epoch": 84.0, "learning_rate": 1.362051282051282e-06, "loss": 0.0, "step": 17350 }, { "epoch": 84.0, "learning_rate": 1.3492307692307693e-06, "loss": 0.0, "step": 17375 }, { "epoch": 84.0, "learning_rate": 1.3364102564102566e-06, "loss": 0.0, "step": 17400 }, { "epoch": 84.01, "learning_rate": 1.3235897435897438e-06, "loss": 0.0, "step": 17425 }, { "epoch": 84.01, "learning_rate": 1.3107692307692308e-06, "loss": 0.0, "step": 17450 }, { "epoch": 84.01, "learning_rate": 1.2979487179487179e-06, "loss": 0.0, "step": 17475 }, { "epoch": 84.01, "learning_rate": 1.2851282051282051e-06, "loss": 0.0, "step": 17500 }, { "epoch": 85.0, "learning_rate": 1.2723076923076924e-06, "loss": 0.0, "step": 17525 }, { "epoch": 85.0, "learning_rate": 1.2594871794871796e-06, "loss": 0.0, "step": 17550 }, { "epoch": 85.0, "learning_rate": 1.2466666666666667e-06, "loss": 0.0, "step": 17575 }, { "epoch": 85.0, "learning_rate": 1.233846153846154e-06, "loss": 0.0, "step": 17600 }, { "epoch": 85.01, "learning_rate": 1.2210256410256412e-06, "loss": 0.0, "step": 17625 }, { "epoch": 85.01, "learning_rate": 1.2082051282051282e-06, "loss": 0.0, "step": 17650 }, { "epoch": 85.01, "learning_rate": 1.1953846153846154e-06, "loss": 0.0, "step": 17675 }, { "epoch": 85.01, "learning_rate": 1.1825641025641027e-06, "loss": 0.0, "step": 17700 }, { "epoch": 86.0, "learning_rate": 1.1697435897435897e-06, "loss": 0.0, "step": 17725 }, { "epoch": 86.0, "learning_rate": 1.156923076923077e-06, "loss": 0.0, "step": 17750 }, { "epoch": 86.0, "learning_rate": 1.1441025641025642e-06, "loss": 0.0, "step": 17775 }, { "epoch": 86.0, "learning_rate": 1.1312820512820513e-06, "loss": 0.0, "step": 17800 }, { "epoch": 86.01, "learning_rate": 1.1184615384615385e-06, "loss": 0.0, "step": 17825 }, { "epoch": 86.01, "learning_rate": 1.1056410256410258e-06, "loss": 0.0, "step": 17850 }, { "epoch": 86.01, "learning_rate": 1.0928205128205128e-06, "loss": 0.0, "step": 17875 }, { "epoch": 86.01, "learning_rate": 1.08e-06, "loss": 0.0, "step": 17900 }, { "epoch": 87.0, "learning_rate": 1.0671794871794873e-06, "loss": 0.0, "step": 17925 }, { "epoch": 87.0, "learning_rate": 1.0543589743589743e-06, "loss": 0.0, "step": 17950 }, { "epoch": 87.0, "learning_rate": 1.0415384615384616e-06, "loss": 0.0, "step": 17975 }, { "epoch": 87.0, "learning_rate": 1.0287179487179488e-06, "loss": 0.0, "step": 18000 }, { "epoch": 87.0, "eval_loss": 0.43187519907951355, "eval_runtime": 2300.0729, "eval_samples_per_second": 2.866, "eval_steps_per_second": 0.179, "eval_wer": 11.861557778767873, "step": 18000 }, { "epoch": 87.01, "learning_rate": 1.0158974358974359e-06, "loss": 0.0, "step": 18025 }, { "epoch": 87.01, "learning_rate": 1.003076923076923e-06, "loss": 0.0, "step": 18050 }, { "epoch": 87.01, "learning_rate": 9.902564102564103e-07, "loss": 0.0, "step": 18075 }, { "epoch": 87.01, "learning_rate": 9.774358974358974e-07, "loss": 0.0, "step": 18100 }, { "epoch": 87.01, "learning_rate": 9.646153846153846e-07, "loss": 0.0, "step": 18125 }, { "epoch": 88.0, "learning_rate": 9.517948717948719e-07, "loss": 0.0, "step": 18150 }, { "epoch": 88.0, "learning_rate": 9.38974358974359e-07, "loss": 0.0, "step": 18175 }, { "epoch": 88.0, "learning_rate": 9.261538461538462e-07, "loss": 0.0, "step": 18200 }, { "epoch": 88.0, "learning_rate": 9.133333333333334e-07, "loss": 0.0, "step": 18225 }, { "epoch": 88.01, "learning_rate": 9.005128205128206e-07, "loss": 0.0, "step": 18250 }, { "epoch": 88.01, "learning_rate": 8.876923076923077e-07, "loss": 0.0, "step": 18275 }, { "epoch": 88.01, "learning_rate": 8.748717948717949e-07, "loss": 0.0, "step": 18300 }, { "epoch": 88.01, "learning_rate": 8.620512820512822e-07, "loss": 0.0, "step": 18325 }, { "epoch": 89.0, "learning_rate": 8.492307692307692e-07, "loss": 0.0, "step": 18350 }, { "epoch": 89.0, "learning_rate": 8.364102564102565e-07, "loss": 0.0, "step": 18375 }, { "epoch": 89.0, "learning_rate": 8.235897435897437e-07, "loss": 0.0, "step": 18400 }, { "epoch": 89.0, "learning_rate": 8.107692307692308e-07, "loss": 0.0, "step": 18425 }, { "epoch": 89.01, "learning_rate": 7.97948717948718e-07, "loss": 0.0, "step": 18450 }, { "epoch": 89.01, "learning_rate": 7.851282051282053e-07, "loss": 0.0, "step": 18475 }, { "epoch": 89.01, "learning_rate": 7.723076923076923e-07, "loss": 0.0, "step": 18500 }, { "epoch": 89.01, "learning_rate": 7.594871794871795e-07, "loss": 0.0, "step": 18525 }, { "epoch": 90.0, "learning_rate": 7.466666666666668e-07, "loss": 0.0, "step": 18550 }, { "epoch": 90.0, "learning_rate": 7.338461538461538e-07, "loss": 0.0, "step": 18575 }, { "epoch": 90.0, "learning_rate": 7.210256410256411e-07, "loss": 0.0, "step": 18600 }, { "epoch": 90.0, "learning_rate": 7.082051282051283e-07, "loss": 0.0, "step": 18625 }, { "epoch": 90.01, "learning_rate": 6.953846153846154e-07, "loss": 0.0, "step": 18650 }, { "epoch": 90.01, "learning_rate": 6.825641025641026e-07, "loss": 0.0, "step": 18675 }, { "epoch": 90.01, "learning_rate": 6.697435897435899e-07, "loss": 0.0, "step": 18700 }, { "epoch": 90.01, "learning_rate": 6.569230769230769e-07, "loss": 0.0, "step": 18725 }, { "epoch": 91.0, "learning_rate": 6.441025641025641e-07, "loss": 0.0, "step": 18750 }, { "epoch": 91.0, "learning_rate": 6.312820512820514e-07, "loss": 0.0, "step": 18775 }, { "epoch": 91.0, "learning_rate": 6.184615384615385e-07, "loss": 0.0, "step": 18800 }, { "epoch": 91.0, "learning_rate": 6.056410256410257e-07, "loss": 0.0, "step": 18825 }, { "epoch": 91.01, "learning_rate": 5.928205128205128e-07, "loss": 0.0, "step": 18850 }, { "epoch": 91.01, "learning_rate": 5.800000000000001e-07, "loss": 0.0, "step": 18875 }, { "epoch": 91.01, "learning_rate": 5.671794871794872e-07, "loss": 0.0, "step": 18900 }, { "epoch": 91.01, "learning_rate": 5.543589743589743e-07, "loss": 0.0, "step": 18925 }, { "epoch": 91.01, "learning_rate": 5.415384615384616e-07, "loss": 0.0, "step": 18950 }, { "epoch": 92.0, "learning_rate": 5.287179487179487e-07, "loss": 0.0, "step": 18975 }, { "epoch": 92.0, "learning_rate": 5.158974358974359e-07, "loss": 0.0, "step": 19000 }, { "epoch": 92.0, "eval_loss": 0.43753913044929504, "eval_runtime": 2323.5793, "eval_samples_per_second": 2.837, "eval_steps_per_second": 0.177, "eval_wer": 11.85345700514399, "step": 19000 }, { "epoch": 92.0, "learning_rate": 5.030769230769231e-07, "loss": 0.0, "step": 19025 }, { "epoch": 92.0, "learning_rate": 4.902564102564103e-07, "loss": 0.0, "step": 19050 }, { "epoch": 92.01, "learning_rate": 4.774358974358974e-07, "loss": 0.0, "step": 19075 }, { "epoch": 92.01, "learning_rate": 4.6461538461538465e-07, "loss": 0.0, "step": 19100 }, { "epoch": 92.01, "learning_rate": 4.5179487179487185e-07, "loss": 0.0, "step": 19125 }, { "epoch": 92.01, "learning_rate": 4.38974358974359e-07, "loss": 0.0, "step": 19150 }, { "epoch": 93.0, "learning_rate": 4.261538461538462e-07, "loss": 0.0, "step": 19175 }, { "epoch": 93.0, "learning_rate": 4.133333333333334e-07, "loss": 0.0, "step": 19200 }, { "epoch": 93.0, "learning_rate": 4.005128205128205e-07, "loss": 0.0, "step": 19225 }, { "epoch": 93.0, "learning_rate": 3.876923076923077e-07, "loss": 0.0, "step": 19250 }, { "epoch": 93.01, "learning_rate": 3.748717948717949e-07, "loss": 0.0, "step": 19275 }, { "epoch": 93.01, "learning_rate": 3.6205128205128206e-07, "loss": 0.0, "step": 19300 }, { "epoch": 93.01, "learning_rate": 3.4923076923076925e-07, "loss": 0.0, "step": 19325 }, { "epoch": 93.01, "learning_rate": 3.3641025641025645e-07, "loss": 0.0, "step": 19350 }, { "epoch": 94.0, "learning_rate": 3.235897435897436e-07, "loss": 0.0, "step": 19375 }, { "epoch": 94.0, "learning_rate": 3.107692307692308e-07, "loss": 0.0, "step": 19400 }, { "epoch": 94.0, "learning_rate": 2.97948717948718e-07, "loss": 0.0, "step": 19425 }, { "epoch": 94.0, "learning_rate": 2.8512820512820517e-07, "loss": 0.0, "step": 19450 }, { "epoch": 94.01, "learning_rate": 2.7230769230769237e-07, "loss": 0.0, "step": 19475 }, { "epoch": 94.01, "learning_rate": 2.594871794871795e-07, "loss": 0.0, "step": 19500 }, { "epoch": 94.01, "learning_rate": 2.466666666666667e-07, "loss": 0.0, "step": 19525 }, { "epoch": 94.01, "learning_rate": 2.3384615384615387e-07, "loss": 0.0, "step": 19550 }, { "epoch": 95.0, "learning_rate": 2.2102564102564104e-07, "loss": 0.0, "step": 19575 }, { "epoch": 95.0, "learning_rate": 2.082051282051282e-07, "loss": 0.0, "step": 19600 }, { "epoch": 95.0, "learning_rate": 1.953846153846154e-07, "loss": 0.0, "step": 19625 }, { "epoch": 95.0, "learning_rate": 1.8256410256410257e-07, "loss": 0.0, "step": 19650 }, { "epoch": 95.01, "learning_rate": 1.6974358974358974e-07, "loss": 0.0, "step": 19675 }, { "epoch": 95.01, "learning_rate": 1.5692307692307694e-07, "loss": 0.0, "step": 19700 }, { "epoch": 95.01, "learning_rate": 1.441025641025641e-07, "loss": 0.0, "step": 19725 }, { "epoch": 95.01, "learning_rate": 1.3128205128205127e-07, "loss": 0.0, "step": 19750 }, { "epoch": 95.01, "learning_rate": 1.1846153846153847e-07, "loss": 0.0, "step": 19775 }, { "epoch": 96.0, "learning_rate": 1.0564102564102565e-07, "loss": 0.0, "step": 19800 }, { "epoch": 96.0, "learning_rate": 9.282051282051283e-08, "loss": 0.0, "step": 19825 }, { "epoch": 96.0, "learning_rate": 8e-08, "loss": 0.0, "step": 19850 }, { "epoch": 96.0, "learning_rate": 6.717948717948718e-08, "loss": 0.0, "step": 19875 }, { "epoch": 96.01, "learning_rate": 5.435897435897436e-08, "loss": 0.0, "step": 19900 }, { "epoch": 96.01, "learning_rate": 4.153846153846154e-08, "loss": 0.0, "step": 19925 }, { "epoch": 96.01, "learning_rate": 2.8717948717948722e-08, "loss": 0.0, "step": 19950 }, { "epoch": 96.01, "learning_rate": 1.5897435897435897e-08, "loss": 0.0, "step": 19975 }, { "epoch": 97.0, "learning_rate": 3.0769230769230774e-09, "loss": 0.0, "step": 20000 }, { "epoch": 97.0, "eval_loss": 0.44003668427467346, "eval_runtime": 2297.3462, "eval_samples_per_second": 2.869, "eval_steps_per_second": 0.179, "eval_wer": 11.865608165579813, "step": 20000 }, { "epoch": 97.0, "step": 20000, "total_flos": 2.7172470114459614e+21, "train_loss": 0.014449456868572451, "train_runtime": 382801.2433, "train_samples_per_second": 3.344, "train_steps_per_second": 0.052 } ], "logging_steps": 25, "max_steps": 20000, "num_train_epochs": 9223372036854775807, "save_steps": 1000, "total_flos": 2.7172470114459614e+21, "trial_name": null, "trial_params": null }