| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 4.1952, | |
| "global_step": 5000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.6000000000000004e-07, | |
| "loss": 1.4968, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.600000000000001e-07, | |
| "loss": 1.3696, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.46e-06, | |
| "loss": 1.2226, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.9600000000000003e-06, | |
| "loss": 1.0319, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.46e-06, | |
| "loss": 0.7679, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.96e-06, | |
| "loss": 0.7524, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3.46e-06, | |
| "loss": 0.6892, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3.96e-06, | |
| "loss": 0.6468, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 4.4600000000000005e-06, | |
| "loss": 0.6165, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.960000000000001e-06, | |
| "loss": 0.6237, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 5.460000000000001e-06, | |
| "loss": 0.6024, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 5.9600000000000005e-06, | |
| "loss": 0.6358, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 6.460000000000001e-06, | |
| "loss": 0.6328, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 6.96e-06, | |
| "loss": 0.583, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 7.4600000000000006e-06, | |
| "loss": 0.579, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 7.960000000000002e-06, | |
| "loss": 0.5774, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 8.46e-06, | |
| "loss": 0.5592, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 8.96e-06, | |
| "loss": 0.552, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.460000000000001e-06, | |
| "loss": 0.5467, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.960000000000001e-06, | |
| "loss": 0.5712, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.94888888888889e-06, | |
| "loss": 0.6092, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.893333333333334e-06, | |
| "loss": 0.5803, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.837777777777778e-06, | |
| "loss": 0.5499, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.782222222222222e-06, | |
| "loss": 0.5281, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.726666666666668e-06, | |
| "loss": 0.5406, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.671111111111112e-06, | |
| "loss": 0.5168, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.615555555555558e-06, | |
| "loss": 0.5312, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.56e-06, | |
| "loss": 0.4867, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.504444444444446e-06, | |
| "loss": 0.4809, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.44888888888889e-06, | |
| "loss": 0.4945, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.393333333333334e-06, | |
| "loss": 0.4777, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.33777777777778e-06, | |
| "loss": 0.5428, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.282222222222222e-06, | |
| "loss": 0.5316, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.226666666666668e-06, | |
| "loss": 0.5616, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.171111111111112e-06, | |
| "loss": 0.5527, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.115555555555556e-06, | |
| "loss": 0.5314, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.060000000000001e-06, | |
| "loss": 0.5178, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.004444444444445e-06, | |
| "loss": 0.5303, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 8.94888888888889e-06, | |
| "loss": 0.5097, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 8.893333333333333e-06, | |
| "loss": 0.534, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "eval_loss": 0.5709909200668335, | |
| "eval_runtime": 3809.9438, | |
| "eval_samples_per_second": 4.223, | |
| "eval_steps_per_second": 0.264, | |
| "eval_wer": 27.440817875610797, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 8.83777777777778e-06, | |
| "loss": 0.5177, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 8.782222222222223e-06, | |
| "loss": 0.4853, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 8.726666666666667e-06, | |
| "loss": 0.4669, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 8.671111111111113e-06, | |
| "loss": 0.4985, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 8.615555555555555e-06, | |
| "loss": 0.4884, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 8.560000000000001e-06, | |
| "loss": 0.4859, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 8.504444444444445e-06, | |
| "loss": 0.4506, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 8.448888888888889e-06, | |
| "loss": 0.4422, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 8.393333333333335e-06, | |
| "loss": 0.4485, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 8.337777777777777e-06, | |
| "loss": 0.4989, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 8.282222222222223e-06, | |
| "loss": 0.4273, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 8.226666666666667e-06, | |
| "loss": 0.4521, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 8.171111111111113e-06, | |
| "loss": 0.4342, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 8.115555555555557e-06, | |
| "loss": 0.4483, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 8.06e-06, | |
| "loss": 0.4358, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 8.004444444444445e-06, | |
| "loss": 0.413, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 7.948888888888889e-06, | |
| "loss": 0.3895, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 7.893333333333335e-06, | |
| "loss": 0.3791, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 7.837777777777779e-06, | |
| "loss": 0.4408, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 7.782222222222223e-06, | |
| "loss": 0.42, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 7.726666666666667e-06, | |
| "loss": 0.4028, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 7.67111111111111e-06, | |
| "loss": 0.4754, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 7.6155555555555564e-06, | |
| "loss": 0.4368, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 7.5600000000000005e-06, | |
| "loss": 0.3854, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 7.504444444444445e-06, | |
| "loss": 0.3895, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 7.44888888888889e-06, | |
| "loss": 0.4029, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 7.393333333333333e-06, | |
| "loss": 0.3991, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 7.337777777777778e-06, | |
| "loss": 0.4328, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 7.282222222222222e-06, | |
| "loss": 0.4165, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 7.226666666666667e-06, | |
| "loss": 0.4338, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 7.171111111111112e-06, | |
| "loss": 0.4157, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 7.115555555555557e-06, | |
| "loss": 0.43, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 7.06e-06, | |
| "loss": 0.4516, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 7.004444444444445e-06, | |
| "loss": 0.4472, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 6.948888888888889e-06, | |
| "loss": 0.4388, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 6.893333333333334e-06, | |
| "loss": 0.414, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 6.837777777777779e-06, | |
| "loss": 0.4528, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 6.782222222222222e-06, | |
| "loss": 0.385, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 6.726666666666667e-06, | |
| "loss": 0.4431, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 6.671111111111112e-06, | |
| "loss": 0.4409, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "eval_loss": 0.5278826355934143, | |
| "eval_runtime": 3790.5651, | |
| "eval_samples_per_second": 4.244, | |
| "eval_steps_per_second": 0.265, | |
| "eval_wer": 25.19806442430855, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 6.615555555555556e-06, | |
| "loss": 0.4096, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 6.560000000000001e-06, | |
| "loss": 0.5584, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 6.504444444444446e-06, | |
| "loss": 0.5934, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 6.448888888888889e-06, | |
| "loss": 0.541, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 6.393333333333334e-06, | |
| "loss": 0.4386, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 6.3377777777777786e-06, | |
| "loss": 0.3835, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 6.282222222222223e-06, | |
| "loss": 0.334, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 6.2266666666666675e-06, | |
| "loss": 0.3392, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 6.171111111111112e-06, | |
| "loss": 0.3311, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 6.1155555555555555e-06, | |
| "loss": 0.3371, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 6.0600000000000004e-06, | |
| "loss": 0.3355, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 6.004444444444445e-06, | |
| "loss": 0.3568, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 5.948888888888889e-06, | |
| "loss": 0.3625, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 5.893333333333334e-06, | |
| "loss": 0.3538, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 5.837777777777777e-06, | |
| "loss": 0.3757, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 5.782222222222222e-06, | |
| "loss": 0.3618, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 5.726666666666667e-06, | |
| "loss": 0.3446, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 5.671111111111112e-06, | |
| "loss": 0.3266, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 5.615555555555556e-06, | |
| "loss": 0.3576, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 5.560000000000001e-06, | |
| "loss": 0.3332, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 5.504444444444444e-06, | |
| "loss": 0.3217, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 5.448888888888889e-06, | |
| "loss": 0.3245, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 5.393333333333334e-06, | |
| "loss": 0.3252, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 5.337777777777779e-06, | |
| "loss": 0.2869, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 5.282222222222223e-06, | |
| "loss": 0.2905, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 5.226666666666667e-06, | |
| "loss": 0.3114, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 5.171111111111111e-06, | |
| "loss": 0.3318, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 5.115555555555556e-06, | |
| "loss": 0.3338, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 5.060000000000001e-06, | |
| "loss": 0.294, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 5.004444444444445e-06, | |
| "loss": 0.3245, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 4.94888888888889e-06, | |
| "loss": 0.3502, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 4.893333333333334e-06, | |
| "loss": 0.3513, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 4.837777777777778e-06, | |
| "loss": 0.3505, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 4.784444444444445e-06, | |
| "loss": 0.3646, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 4.728888888888889e-06, | |
| "loss": 0.3493, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 4.673333333333333e-06, | |
| "loss": 0.3257, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 4.617777777777778e-06, | |
| "loss": 0.3432, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 4.562222222222222e-06, | |
| "loss": 0.3238, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 4.506666666666667e-06, | |
| "loss": 0.2973, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 4.451111111111112e-06, | |
| "loss": 0.3095, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "eval_loss": 0.5117268562316895, | |
| "eval_runtime": 3800.0871, | |
| "eval_samples_per_second": 4.234, | |
| "eval_steps_per_second": 0.265, | |
| "eval_wer": 25.08183500166042, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 4.395555555555556e-06, | |
| "loss": 0.3395, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 4.34e-06, | |
| "loss": 0.3365, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 4.284444444444445e-06, | |
| "loss": 0.3727, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 4.228888888888889e-06, | |
| "loss": 0.3507, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 4.173333333333334e-06, | |
| "loss": 0.3635, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 4.117777777777779e-06, | |
| "loss": 0.3309, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 4.062222222222223e-06, | |
| "loss": 0.3317, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 4.006666666666667e-06, | |
| "loss": 0.3199, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 3.951111111111112e-06, | |
| "loss": 0.3149, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 3.895555555555556e-06, | |
| "loss": 0.3387, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 3.8400000000000005e-06, | |
| "loss": 0.307, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 3.784444444444445e-06, | |
| "loss": 0.3933, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 3.728888888888889e-06, | |
| "loss": 0.3645, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 3.673333333333334e-06, | |
| "loss": 0.334, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 3.617777777777778e-06, | |
| "loss": 0.3325, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 3.5622222222222224e-06, | |
| "loss": 0.3125, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 3.5066666666666673e-06, | |
| "loss": 0.2657, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 3.4511111111111113e-06, | |
| "loss": 0.3109, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 3.3955555555555558e-06, | |
| "loss": 0.2904, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 3.3400000000000006e-06, | |
| "loss": 0.2891, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 3.2844444444444447e-06, | |
| "loss": 0.3126, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 3.228888888888889e-06, | |
| "loss": 0.3429, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 3.173333333333334e-06, | |
| "loss": 0.3201, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 3.117777777777778e-06, | |
| "loss": 0.3232, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 3.0622222222222225e-06, | |
| "loss": 0.3105, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 3.0066666666666674e-06, | |
| "loss": 0.3151, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 2.9511111111111114e-06, | |
| "loss": 0.3098, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 2.895555555555556e-06, | |
| "loss": 0.3121, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 2.84e-06, | |
| "loss": 0.3173, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 2.784444444444445e-06, | |
| "loss": 0.3034, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 2.7288888888888893e-06, | |
| "loss": 0.3064, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 2.6733333333333333e-06, | |
| "loss": 0.2941, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 2.617777777777778e-06, | |
| "loss": 0.3511, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 2.5622222222222226e-06, | |
| "loss": 0.3365, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 2.5066666666666667e-06, | |
| "loss": 0.3109, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 2.451111111111111e-06, | |
| "loss": 0.3242, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 2.3955555555555556e-06, | |
| "loss": 0.3324, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 2.3400000000000005e-06, | |
| "loss": 0.3344, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 2.2844444444444445e-06, | |
| "loss": 0.3154, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 2.228888888888889e-06, | |
| "loss": 0.3285, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "eval_loss": 0.49949249625205994, | |
| "eval_runtime": 3794.7248, | |
| "eval_samples_per_second": 4.24, | |
| "eval_steps_per_second": 0.265, | |
| "eval_wer": 24.06008349542198, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 2.1733333333333334e-06, | |
| "loss": 0.3334, | |
| "step": 4025 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 2.117777777777778e-06, | |
| "loss": 0.3189, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 2.0622222222222223e-06, | |
| "loss": 0.3228, | |
| "step": 4075 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 2.008888888888889e-06, | |
| "loss": 0.3635, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 1.9533333333333334e-06, | |
| "loss": 0.3541, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 1.8977777777777779e-06, | |
| "loss": 0.3155, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 1.8422222222222225e-06, | |
| "loss": 0.3045, | |
| "step": 4175 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 1.7866666666666668e-06, | |
| "loss": 0.3317, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 1.7311111111111112e-06, | |
| "loss": 0.3121, | |
| "step": 4225 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 1.675555555555556e-06, | |
| "loss": 0.3118, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 1.6200000000000002e-06, | |
| "loss": 0.3039, | |
| "step": 4275 | |
| }, | |
| { | |
| "epoch": 4.06, | |
| "learning_rate": 1.5644444444444446e-06, | |
| "loss": 0.3756, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 4.06, | |
| "learning_rate": 1.5088888888888889e-06, | |
| "loss": 0.3453, | |
| "step": 4325 | |
| }, | |
| { | |
| "epoch": 4.07, | |
| "learning_rate": 1.4533333333333335e-06, | |
| "loss": 0.3875, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 4.07, | |
| "learning_rate": 1.397777777777778e-06, | |
| "loss": 0.3773, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "learning_rate": 1.3422222222222222e-06, | |
| "loss": 0.3819, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "learning_rate": 1.286666666666667e-06, | |
| "loss": 0.357, | |
| "step": 4425 | |
| }, | |
| { | |
| "epoch": 4.09, | |
| "learning_rate": 1.2311111111111112e-06, | |
| "loss": 0.3456, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 4.09, | |
| "learning_rate": 1.1755555555555556e-06, | |
| "loss": 0.3685, | |
| "step": 4475 | |
| }, | |
| { | |
| "epoch": 4.1, | |
| "learning_rate": 1.12e-06, | |
| "loss": 0.3523, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 4.1, | |
| "learning_rate": 1.0644444444444445e-06, | |
| "loss": 0.3175, | |
| "step": 4525 | |
| }, | |
| { | |
| "epoch": 4.11, | |
| "learning_rate": 1.008888888888889e-06, | |
| "loss": 0.319, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 4.11, | |
| "learning_rate": 9.533333333333335e-07, | |
| "loss": 0.2706, | |
| "step": 4575 | |
| }, | |
| { | |
| "epoch": 4.12, | |
| "learning_rate": 8.977777777777778e-07, | |
| "loss": 0.2225, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 4.12, | |
| "learning_rate": 8.422222222222224e-07, | |
| "loss": 0.2467, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 4.13, | |
| "learning_rate": 7.866666666666667e-07, | |
| "loss": 0.2587, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 4.13, | |
| "learning_rate": 7.311111111111112e-07, | |
| "loss": 0.2409, | |
| "step": 4675 | |
| }, | |
| { | |
| "epoch": 4.14, | |
| "learning_rate": 6.755555555555555e-07, | |
| "loss": 0.2418, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 4.14, | |
| "learning_rate": 6.200000000000001e-07, | |
| "loss": 0.2309, | |
| "step": 4725 | |
| }, | |
| { | |
| "epoch": 4.15, | |
| "learning_rate": 5.644444444444445e-07, | |
| "loss": 0.2666, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 4.15, | |
| "learning_rate": 5.088888888888889e-07, | |
| "loss": 0.2854, | |
| "step": 4775 | |
| }, | |
| { | |
| "epoch": 4.16, | |
| "learning_rate": 4.533333333333334e-07, | |
| "loss": 0.2987, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 4.16, | |
| "learning_rate": 3.9777777777777783e-07, | |
| "loss": 0.2957, | |
| "step": 4825 | |
| }, | |
| { | |
| "epoch": 4.17, | |
| "learning_rate": 3.422222222222223e-07, | |
| "loss": 0.3315, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 4.17, | |
| "learning_rate": 2.866666666666667e-07, | |
| "loss": 0.2915, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 4.18, | |
| "learning_rate": 2.3111111111111112e-07, | |
| "loss": 0.3233, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 4.18, | |
| "learning_rate": 1.7555555555555558e-07, | |
| "loss": 0.3344, | |
| "step": 4925 | |
| }, | |
| { | |
| "epoch": 4.19, | |
| "learning_rate": 1.2000000000000002e-07, | |
| "loss": 0.3206, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 4.19, | |
| "learning_rate": 6.444444444444445e-08, | |
| "loss": 0.2938, | |
| "step": 4975 | |
| }, | |
| { | |
| "epoch": 4.2, | |
| "learning_rate": 8.88888888888889e-09, | |
| "loss": 0.3032, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 4.2, | |
| "eval_loss": 0.496755987405777, | |
| "eval_runtime": 3794.9348, | |
| "eval_samples_per_second": 4.24, | |
| "eval_steps_per_second": 0.265, | |
| "eval_wer": 24.064827553489256, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 4.2, | |
| "step": 5000, | |
| "total_flos": 1.037396722581504e+19, | |
| "train_loss": 0.41511208391189575, | |
| "train_runtime": 60945.1133, | |
| "train_samples_per_second": 2.625, | |
| "train_steps_per_second": 0.082 | |
| } | |
| ], | |
| "max_steps": 5000, | |
| "num_train_epochs": 9223372036854775807, | |
| "total_flos": 1.037396722581504e+19, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |