| { | |
| "best_metric": 20.468053491827636, | |
| "best_model_checkpoint": "./data/finetuningRuns/whisper-sm-el-intlv-xl/checkpoint-5000", | |
| "epoch": 12.437810945273633, | |
| "global_step": 5000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 3.0000000000000004e-07, | |
| "loss": 0.9128, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 6.125000000000001e-07, | |
| "loss": 0.868, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.25e-07, | |
| "loss": 0.7182, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.2375000000000002e-06, | |
| "loss": 0.5856, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.55e-06, | |
| "loss": 0.4819, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.8625e-06, | |
| "loss": 0.4141, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 2.175e-06, | |
| "loss": 0.3834, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.4875000000000003e-06, | |
| "loss": 0.3549, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.8000000000000003e-06, | |
| "loss": 0.3543, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.1125000000000003e-06, | |
| "loss": 0.3167, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 3.4250000000000002e-06, | |
| "loss": 0.3044, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.7375000000000002e-06, | |
| "loss": 0.2813, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 4.05e-06, | |
| "loss": 0.2801, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 4.3625e-06, | |
| "loss": 0.2573, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 4.675e-06, | |
| "loss": 0.2407, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 4.9875e-06, | |
| "loss": 0.2459, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 5.3e-06, | |
| "loss": 0.2009, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 5.6125e-06, | |
| "loss": 0.1957, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 5.925e-06, | |
| "loss": 0.1779, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 6.2375e-06, | |
| "loss": 0.1798, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 6.2166666666666676e-06, | |
| "loss": 0.1636, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 6.181944444444445e-06, | |
| "loss": 0.1698, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 6.147222222222222e-06, | |
| "loss": 0.17, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 6.1125e-06, | |
| "loss": 0.1737, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 6.077777777777778e-06, | |
| "loss": 0.1664, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 6.043055555555556e-06, | |
| "loss": 0.1459, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 6.008333333333334e-06, | |
| "loss": 0.1457, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 5.973611111111111e-06, | |
| "loss": 0.1571, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 5.938888888888889e-06, | |
| "loss": 0.1337, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 5.904166666666667e-06, | |
| "loss": 0.1337, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 5.869444444444445e-06, | |
| "loss": 0.1464, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 5.834722222222223e-06, | |
| "loss": 0.1354, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 5.8e-06, | |
| "loss": 0.0992, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 5.765277777777778e-06, | |
| "loss": 0.0791, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 5.730555555555556e-06, | |
| "loss": 0.0742, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 5.695833333333334e-06, | |
| "loss": 0.0827, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 5.661111111111111e-06, | |
| "loss": 0.0844, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 5.6263888888888895e-06, | |
| "loss": 0.0867, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 5.591666666666668e-06, | |
| "loss": 0.0699, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 5.556944444444444e-06, | |
| "loss": 0.0705, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "eval_loss": 0.2869577705860138, | |
| "eval_runtime": 1144.7947, | |
| "eval_samples_per_second": 1.481, | |
| "eval_steps_per_second": 0.185, | |
| "eval_wer": 21.49888558692422, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 5.522222222222222e-06, | |
| "loss": 0.0684, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 5.4875e-06, | |
| "loss": 0.0833, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 5.4527777777777786e-06, | |
| "loss": 0.0737, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 5.418055555555556e-06, | |
| "loss": 0.0748, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 5.383333333333333e-06, | |
| "loss": 0.0672, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 5.348611111111111e-06, | |
| "loss": 0.0741, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 5.3138888888888895e-06, | |
| "loss": 0.0683, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 5.279166666666667e-06, | |
| "loss": 0.0665, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 5.244444444444445e-06, | |
| "loss": 0.0449, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 5.209722222222223e-06, | |
| "loss": 0.0362, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 5.175e-06, | |
| "loss": 0.0377, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 5.140277777777778e-06, | |
| "loss": 0.0351, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 5.105555555555556e-06, | |
| "loss": 0.037, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 5.070833333333334e-06, | |
| "loss": 0.0363, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 5.036111111111111e-06, | |
| "loss": 0.0324, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 5.001388888888889e-06, | |
| "loss": 0.0344, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 4.966666666666667e-06, | |
| "loss": 0.04, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 3.61, | |
| "learning_rate": 4.931944444444445e-06, | |
| "loss": 0.0369, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 4.897222222222222e-06, | |
| "loss": 0.0369, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 4.8625000000000005e-06, | |
| "loss": 0.0344, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 4.827777777777779e-06, | |
| "loss": 0.0311, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 4.793055555555555e-06, | |
| "loss": 0.0296, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 4.758333333333333e-06, | |
| "loss": 0.0333, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 4.723611111111111e-06, | |
| "loss": 0.0322, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 4.6888888888888895e-06, | |
| "loss": 0.0216, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 4.1, | |
| "learning_rate": 4.654166666666667e-06, | |
| "loss": 0.0139, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 4.17, | |
| "learning_rate": 4.619444444444445e-06, | |
| "loss": 0.015, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 4.23, | |
| "learning_rate": 4.584722222222222e-06, | |
| "loss": 0.0184, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 4.29, | |
| "learning_rate": 4.5500000000000005e-06, | |
| "loss": 0.014, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 4.35, | |
| "learning_rate": 4.515277777777778e-06, | |
| "loss": 0.0166, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 4.42, | |
| "learning_rate": 4.480555555555556e-06, | |
| "loss": 0.0144, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 4.48, | |
| "learning_rate": 4.445833333333334e-06, | |
| "loss": 0.0159, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 4.54, | |
| "learning_rate": 4.411111111111111e-06, | |
| "loss": 0.0148, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 4.6, | |
| "learning_rate": 4.376388888888889e-06, | |
| "loss": 0.0155, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 4.66, | |
| "learning_rate": 4.341666666666667e-06, | |
| "loss": 0.0137, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 4.73, | |
| "learning_rate": 4.306944444444445e-06, | |
| "loss": 0.0158, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 4.79, | |
| "learning_rate": 4.272222222222222e-06, | |
| "loss": 0.0133, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 4.85, | |
| "learning_rate": 4.2375000000000005e-06, | |
| "loss": 0.0149, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 4.91, | |
| "learning_rate": 4.202777777777778e-06, | |
| "loss": 0.0147, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "learning_rate": 4.168055555555556e-06, | |
| "loss": 0.0147, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "eval_loss": 0.3688688576221466, | |
| "eval_runtime": 1052.2424, | |
| "eval_samples_per_second": 1.612, | |
| "eval_steps_per_second": 0.201, | |
| "eval_wer": 21.01597325408618, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 4.133333333333333e-06, | |
| "loss": 0.0106, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "learning_rate": 4.0986111111111114e-06, | |
| "loss": 0.0072, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 5.16, | |
| "learning_rate": 4.06388888888889e-06, | |
| "loss": 0.0084, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 5.22, | |
| "learning_rate": 4.029166666666667e-06, | |
| "loss": 0.0061, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 5.29, | |
| "learning_rate": 3.994444444444444e-06, | |
| "loss": 0.007, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 5.35, | |
| "learning_rate": 3.959722222222222e-06, | |
| "loss": 0.0075, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 5.41, | |
| "learning_rate": 3.9250000000000005e-06, | |
| "loss": 0.0067, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 5.47, | |
| "learning_rate": 3.890277777777778e-06, | |
| "loss": 0.0065, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 5.53, | |
| "learning_rate": 3.855555555555556e-06, | |
| "loss": 0.0072, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 5.6, | |
| "learning_rate": 3.820833333333333e-06, | |
| "loss": 0.0064, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 5.66, | |
| "learning_rate": 3.786111111111111e-06, | |
| "loss": 0.0072, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 5.72, | |
| "learning_rate": 3.751388888888889e-06, | |
| "loss": 0.0059, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 5.78, | |
| "learning_rate": 3.716666666666667e-06, | |
| "loss": 0.0072, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 5.85, | |
| "learning_rate": 3.6819444444444447e-06, | |
| "loss": 0.0061, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 5.91, | |
| "learning_rate": 3.647222222222223e-06, | |
| "loss": 0.0089, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 5.97, | |
| "learning_rate": 3.6124999999999997e-06, | |
| "loss": 0.0061, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 3.577777777777778e-06, | |
| "loss": 0.0057, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 6.09, | |
| "learning_rate": 3.5430555555555556e-06, | |
| "loss": 0.0044, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 6.16, | |
| "learning_rate": 3.5083333333333338e-06, | |
| "loss": 0.0043, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 6.22, | |
| "learning_rate": 3.4736111111111115e-06, | |
| "loss": 0.0038, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 6.28, | |
| "learning_rate": 3.438888888888889e-06, | |
| "loss": 0.0043, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 6.34, | |
| "learning_rate": 3.4041666666666665e-06, | |
| "loss": 0.0036, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 6.41, | |
| "learning_rate": 3.3694444444444447e-06, | |
| "loss": 0.0038, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 6.47, | |
| "learning_rate": 3.3347222222222224e-06, | |
| "loss": 0.0033, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 6.53, | |
| "learning_rate": 3.3e-06, | |
| "loss": 0.0047, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 6.59, | |
| "learning_rate": 3.2652777777777783e-06, | |
| "loss": 0.0036, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 6.65, | |
| "learning_rate": 3.2305555555555552e-06, | |
| "loss": 0.0034, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 6.72, | |
| "learning_rate": 3.1958333333333334e-06, | |
| "loss": 0.0034, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 6.78, | |
| "learning_rate": 3.161111111111111e-06, | |
| "loss": 0.0036, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 6.84, | |
| "learning_rate": 3.1263888888888893e-06, | |
| "loss": 0.0032, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 6.9, | |
| "learning_rate": 3.0916666666666666e-06, | |
| "loss": 0.0036, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 6.97, | |
| "learning_rate": 3.0569444444444447e-06, | |
| "loss": 0.0031, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 3.0222222222222225e-06, | |
| "loss": 0.0028, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 7.09, | |
| "learning_rate": 2.9875e-06, | |
| "loss": 0.0026, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 7.15, | |
| "learning_rate": 2.952777777777778e-06, | |
| "loss": 0.0029, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 7.21, | |
| "learning_rate": 2.9180555555555557e-06, | |
| "loss": 0.0027, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 7.28, | |
| "learning_rate": 2.8833333333333334e-06, | |
| "loss": 0.0029, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 7.34, | |
| "learning_rate": 2.848611111111111e-06, | |
| "loss": 0.0024, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 7.4, | |
| "learning_rate": 2.8138888888888893e-06, | |
| "loss": 0.0025, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 7.46, | |
| "learning_rate": 2.7791666666666666e-06, | |
| "loss": 0.0024, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 7.46, | |
| "eval_loss": 0.4156392514705658, | |
| "eval_runtime": 1204.6242, | |
| "eval_samples_per_second": 1.408, | |
| "eval_steps_per_second": 0.176, | |
| "eval_wer": 20.75594353640416, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 7.52, | |
| "learning_rate": 2.7444444444444448e-06, | |
| "loss": 0.0026, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 7.59, | |
| "learning_rate": 2.7097222222222225e-06, | |
| "loss": 0.0021, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 7.65, | |
| "learning_rate": 2.6750000000000002e-06, | |
| "loss": 0.0023, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 7.71, | |
| "learning_rate": 2.640277777777778e-06, | |
| "loss": 0.0022, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 7.77, | |
| "learning_rate": 2.6055555555555557e-06, | |
| "loss": 0.0024, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 7.84, | |
| "learning_rate": 2.5708333333333334e-06, | |
| "loss": 0.0027, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 7.9, | |
| "learning_rate": 2.536111111111111e-06, | |
| "loss": 0.0021, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 7.96, | |
| "learning_rate": 2.501388888888889e-06, | |
| "loss": 0.0023, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 2.4666666666666666e-06, | |
| "loss": 0.0019, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 8.08, | |
| "learning_rate": 2.4319444444444448e-06, | |
| "loss": 0.0018, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 8.15, | |
| "learning_rate": 2.397222222222222e-06, | |
| "loss": 0.0018, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 8.21, | |
| "learning_rate": 2.3625000000000003e-06, | |
| "loss": 0.0018, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 8.27, | |
| "learning_rate": 2.327777777777778e-06, | |
| "loss": 0.0021, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 8.33, | |
| "learning_rate": 2.2930555555555557e-06, | |
| "loss": 0.0017, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 8.4, | |
| "learning_rate": 2.2583333333333335e-06, | |
| "loss": 0.0017, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 8.46, | |
| "learning_rate": 2.223611111111111e-06, | |
| "loss": 0.0018, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 8.52, | |
| "learning_rate": 2.188888888888889e-06, | |
| "loss": 0.0017, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 8.58, | |
| "learning_rate": 2.1541666666666667e-06, | |
| "loss": 0.0019, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 8.64, | |
| "learning_rate": 2.1194444444444444e-06, | |
| "loss": 0.0017, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 8.71, | |
| "learning_rate": 2.084722222222222e-06, | |
| "loss": 0.0017, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 8.77, | |
| "learning_rate": 2.0500000000000003e-06, | |
| "loss": 0.0016, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 8.83, | |
| "learning_rate": 2.0152777777777776e-06, | |
| "loss": 0.0019, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 8.89, | |
| "learning_rate": 1.9805555555555557e-06, | |
| "loss": 0.0017, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 8.96, | |
| "learning_rate": 1.9458333333333335e-06, | |
| "loss": 0.0016, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 9.02, | |
| "learning_rate": 1.9111111111111112e-06, | |
| "loss": 0.0016, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 9.08, | |
| "learning_rate": 1.876388888888889e-06, | |
| "loss": 0.0015, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 9.14, | |
| "learning_rate": 1.8416666666666669e-06, | |
| "loss": 0.0016, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 9.2, | |
| "learning_rate": 1.8069444444444444e-06, | |
| "loss": 0.0015, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 9.27, | |
| "learning_rate": 1.7722222222222224e-06, | |
| "loss": 0.0014, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 9.33, | |
| "learning_rate": 1.7375000000000003e-06, | |
| "loss": 0.0015, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 9.39, | |
| "learning_rate": 1.7027777777777778e-06, | |
| "loss": 0.0014, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 9.45, | |
| "learning_rate": 1.6680555555555558e-06, | |
| "loss": 0.0014, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 9.51, | |
| "learning_rate": 1.6333333333333333e-06, | |
| "loss": 0.0014, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 9.58, | |
| "learning_rate": 1.5986111111111112e-06, | |
| "loss": 0.0018, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 9.64, | |
| "learning_rate": 1.563888888888889e-06, | |
| "loss": 0.0014, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 9.7, | |
| "learning_rate": 1.5291666666666667e-06, | |
| "loss": 0.0014, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 9.76, | |
| "learning_rate": 1.4944444444444444e-06, | |
| "loss": 0.0014, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 9.83, | |
| "learning_rate": 1.4597222222222222e-06, | |
| "loss": 0.0014, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 9.89, | |
| "learning_rate": 1.4250000000000001e-06, | |
| "loss": 0.0014, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 9.95, | |
| "learning_rate": 1.3902777777777779e-06, | |
| "loss": 0.0014, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 9.95, | |
| "eval_loss": 0.4423137605190277, | |
| "eval_runtime": 1022.3489, | |
| "eval_samples_per_second": 1.659, | |
| "eval_steps_per_second": 0.207, | |
| "eval_wer": 20.700222882615154, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 10.01, | |
| "learning_rate": 1.3555555555555556e-06, | |
| "loss": 0.0013, | |
| "step": 4025 | |
| }, | |
| { | |
| "epoch": 10.07, | |
| "learning_rate": 1.3208333333333335e-06, | |
| "loss": 0.0013, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 10.14, | |
| "learning_rate": 1.2861111111111113e-06, | |
| "loss": 0.0012, | |
| "step": 4075 | |
| }, | |
| { | |
| "epoch": 10.2, | |
| "learning_rate": 1.251388888888889e-06, | |
| "loss": 0.0013, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 10.26, | |
| "learning_rate": 1.2166666666666667e-06, | |
| "loss": 0.0013, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 10.32, | |
| "learning_rate": 1.1819444444444445e-06, | |
| "loss": 0.0012, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 10.39, | |
| "learning_rate": 1.1472222222222222e-06, | |
| "loss": 0.0012, | |
| "step": 4175 | |
| }, | |
| { | |
| "epoch": 10.45, | |
| "learning_rate": 1.1125e-06, | |
| "loss": 0.0011, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 10.51, | |
| "learning_rate": 1.0777777777777779e-06, | |
| "loss": 0.0012, | |
| "step": 4225 | |
| }, | |
| { | |
| "epoch": 10.57, | |
| "learning_rate": 1.0430555555555556e-06, | |
| "loss": 0.0012, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 10.63, | |
| "learning_rate": 1.0083333333333333e-06, | |
| "loss": 0.0012, | |
| "step": 4275 | |
| }, | |
| { | |
| "epoch": 10.7, | |
| "learning_rate": 9.73611111111111e-07, | |
| "loss": 0.0013, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 10.76, | |
| "learning_rate": 9.38888888888889e-07, | |
| "loss": 0.0012, | |
| "step": 4325 | |
| }, | |
| { | |
| "epoch": 10.82, | |
| "learning_rate": 9.041666666666667e-07, | |
| "loss": 0.0013, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 10.88, | |
| "learning_rate": 8.694444444444444e-07, | |
| "loss": 0.0012, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 10.95, | |
| "learning_rate": 8.347222222222223e-07, | |
| "loss": 0.0012, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 11.01, | |
| "learning_rate": 8.000000000000001e-07, | |
| "loss": 0.0012, | |
| "step": 4425 | |
| }, | |
| { | |
| "epoch": 11.07, | |
| "learning_rate": 7.652777777777778e-07, | |
| "loss": 0.0012, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 11.13, | |
| "learning_rate": 7.305555555555556e-07, | |
| "loss": 0.0011, | |
| "step": 4475 | |
| }, | |
| { | |
| "epoch": 11.19, | |
| "learning_rate": 6.958333333333334e-07, | |
| "loss": 0.001, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 11.26, | |
| "learning_rate": 6.611111111111111e-07, | |
| "loss": 0.0011, | |
| "step": 4525 | |
| }, | |
| { | |
| "epoch": 11.32, | |
| "learning_rate": 6.26388888888889e-07, | |
| "loss": 0.0012, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 11.38, | |
| "learning_rate": 5.916666666666667e-07, | |
| "loss": 0.0011, | |
| "step": 4575 | |
| }, | |
| { | |
| "epoch": 11.44, | |
| "learning_rate": 5.569444444444445e-07, | |
| "loss": 0.0012, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 11.5, | |
| "learning_rate": 5.222222222222223e-07, | |
| "loss": 0.0011, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 11.57, | |
| "learning_rate": 4.875e-07, | |
| "loss": 0.0011, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 11.63, | |
| "learning_rate": 4.5277777777777783e-07, | |
| "loss": 0.0011, | |
| "step": 4675 | |
| }, | |
| { | |
| "epoch": 11.69, | |
| "learning_rate": 4.1805555555555556e-07, | |
| "loss": 0.0012, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 11.75, | |
| "learning_rate": 3.8333333333333335e-07, | |
| "loss": 0.0011, | |
| "step": 4725 | |
| }, | |
| { | |
| "epoch": 11.82, | |
| "learning_rate": 3.4861111111111114e-07, | |
| "loss": 0.0011, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 11.88, | |
| "learning_rate": 3.138888888888889e-07, | |
| "loss": 0.0011, | |
| "step": 4775 | |
| }, | |
| { | |
| "epoch": 11.94, | |
| "learning_rate": 2.7916666666666666e-07, | |
| "loss": 0.0011, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "learning_rate": 2.4444444444444445e-07, | |
| "loss": 0.001, | |
| "step": 4825 | |
| }, | |
| { | |
| "epoch": 12.06, | |
| "learning_rate": 2.0972222222222223e-07, | |
| "loss": 0.0011, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 12.13, | |
| "learning_rate": 1.7500000000000002e-07, | |
| "loss": 0.001, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 12.19, | |
| "learning_rate": 1.4027777777777778e-07, | |
| "loss": 0.001, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 12.25, | |
| "learning_rate": 1.0555555555555555e-07, | |
| "loss": 0.001, | |
| "step": 4925 | |
| }, | |
| { | |
| "epoch": 12.31, | |
| "learning_rate": 7.083333333333334e-08, | |
| "loss": 0.0011, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 12.38, | |
| "learning_rate": 3.611111111111111e-08, | |
| "loss": 0.0012, | |
| "step": 4975 | |
| }, | |
| { | |
| "epoch": 12.44, | |
| "learning_rate": 1.388888888888889e-09, | |
| "loss": 0.0011, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 12.44, | |
| "eval_loss": 0.4527539908885956, | |
| "eval_runtime": 1125.4439, | |
| "eval_samples_per_second": 1.507, | |
| "eval_steps_per_second": 0.188, | |
| "eval_wer": 20.468053491827636, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 12.44, | |
| "step": 5000, | |
| "total_flos": 4.615288590237696e+19, | |
| "train_loss": 0.059744430732727054, | |
| "train_runtime": 38974.8058, | |
| "train_samples_per_second": 4.105, | |
| "train_steps_per_second": 0.128 | |
| } | |
| ], | |
| "max_steps": 5000, | |
| "num_train_epochs": 13, | |
| "total_flos": 4.615288590237696e+19, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |