| { | |
| "best_metric": 91.09343588847129, | |
| "best_model_checkpoint": "./checkpoint-4000", | |
| "epoch": 11.0012, | |
| "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": 4.0000000000000003e-07, | |
| "loss": 1.8238, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.000000000000001e-07, | |
| "loss": 1.4975, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.4000000000000001e-06, | |
| "loss": 1.5217, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.9000000000000002e-06, | |
| "loss": 1.269, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.4000000000000003e-06, | |
| "loss": 1.1025, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.9e-06, | |
| "loss": 0.9731, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3.4000000000000005e-06, | |
| "loss": 0.9875, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3.900000000000001e-06, | |
| "loss": 1.0671, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 4.4e-06, | |
| "loss": 1.039, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.9000000000000005e-06, | |
| "loss": 0.9246, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 5.400000000000001e-06, | |
| "loss": 0.814, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 5.9e-06, | |
| "loss": 0.7673, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 6.4000000000000006e-06, | |
| "loss": 0.6654, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 6.9e-06, | |
| "loss": 0.7185, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 7.4e-06, | |
| "loss": 0.6773, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 7.9e-06, | |
| "loss": 0.6931, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 8.400000000000001e-06, | |
| "loss": 0.6353, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 8.900000000000001e-06, | |
| "loss": 0.5768, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 9.4e-06, | |
| "loss": 1.0683, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 9.9e-06, | |
| "loss": 1.0733, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 9.955555555555556e-06, | |
| "loss": 1.0536, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 9.9e-06, | |
| "loss": 1.0223, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 9.844444444444446e-06, | |
| "loss": 0.856, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 9.78888888888889e-06, | |
| "loss": 0.7194, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 9.733333333333334e-06, | |
| "loss": 0.7801, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 9.677777777777778e-06, | |
| "loss": 0.8352, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 9.622222222222222e-06, | |
| "loss": 0.792, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 9.566666666666668e-06, | |
| "loss": 0.7702, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 9.511111111111112e-06, | |
| "loss": 0.6291, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 9.455555555555557e-06, | |
| "loss": 0.6016, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 9.4e-06, | |
| "loss": 0.5448, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 9.344444444444446e-06, | |
| "loss": 0.5598, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 9.28888888888889e-06, | |
| "loss": 0.5188, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 9.233333333333334e-06, | |
| "loss": 0.5386, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 9.17777777777778e-06, | |
| "loss": 0.5219, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 9.122222222222223e-06, | |
| "loss": 0.4535, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 9.066666666666667e-06, | |
| "loss": 0.826, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 9.011111111111111e-06, | |
| "loss": 0.9529, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 8.955555555555555e-06, | |
| "loss": 0.8913, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 8.900000000000001e-06, | |
| "loss": 0.9397, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "eval_loss": 0.656823992729187, | |
| "eval_runtime": 1612.4976, | |
| "eval_samples_per_second": 6.562, | |
| "eval_steps_per_second": 0.205, | |
| "eval_wer": 98.7326282667599, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 8.844444444444445e-06, | |
| "loss": 0.7235, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 8.788888888888891e-06, | |
| "loss": 0.6305, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 8.733333333333333e-06, | |
| "loss": 0.6316, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 8.677777777777779e-06, | |
| "loss": 0.7288, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 8.622222222222223e-06, | |
| "loss": 0.6636, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 8.566666666666667e-06, | |
| "loss": 0.6748, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 8.511111111111113e-06, | |
| "loss": 0.5362, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 8.455555555555555e-06, | |
| "loss": 0.5111, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 8.400000000000001e-06, | |
| "loss": 0.4571, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 8.344444444444445e-06, | |
| "loss": 0.4781, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 8.288888888888889e-06, | |
| "loss": 0.4331, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 8.233333333333335e-06, | |
| "loss": 0.4486, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 8.177777777777779e-06, | |
| "loss": 0.4374, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 8.122222222222223e-06, | |
| "loss": 0.3948, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 8.066666666666667e-06, | |
| "loss": 0.6683, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 8.011111111111113e-06, | |
| "loss": 0.9137, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 7.955555555555557e-06, | |
| "loss": 0.7532, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 7.9e-06, | |
| "loss": 0.9076, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 7.844444444444446e-06, | |
| "loss": 0.6435, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 7.788888888888889e-06, | |
| "loss": 0.5592, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 7.733333333333334e-06, | |
| "loss": 0.5847, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 7.677777777777778e-06, | |
| "loss": 0.6266, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 7.622222222222223e-06, | |
| "loss": 0.5801, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 7.566666666666667e-06, | |
| "loss": 0.6152, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 7.511111111111111e-06, | |
| "loss": 0.4823, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 7.455555555555556e-06, | |
| "loss": 0.4425, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 7.4e-06, | |
| "loss": 0.4237, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 7.344444444444445e-06, | |
| "loss": 0.4056, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 7.28888888888889e-06, | |
| "loss": 0.3723, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 7.233333333333334e-06, | |
| "loss": 0.3823, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 7.177777777777778e-06, | |
| "loss": 0.3906, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 7.122222222222222e-06, | |
| "loss": 0.3569, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 7.066666666666667e-06, | |
| "loss": 0.5093, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 7.011111111111112e-06, | |
| "loss": 0.9007, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 6.955555555555557e-06, | |
| "loss": 0.6885, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 6.9e-06, | |
| "loss": 0.8392, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 6.844444444444445e-06, | |
| "loss": 0.6147, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 6.788888888888889e-06, | |
| "loss": 0.5308, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 6.733333333333334e-06, | |
| "loss": 0.5261, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 6.677777777777779e-06, | |
| "loss": 0.5387, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "eval_loss": 0.6149321794509888, | |
| "eval_runtime": 1515.2771, | |
| "eval_samples_per_second": 6.983, | |
| "eval_steps_per_second": 0.218, | |
| "eval_wer": 94.51970981557555, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 6.6222222222222236e-06, | |
| "loss": 0.5447, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 6.566666666666667e-06, | |
| "loss": 0.5586, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 6.511111111111112e-06, | |
| "loss": 0.4592, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 4.06, | |
| "learning_rate": 6.455555555555556e-06, | |
| "loss": 0.3925, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 4.06, | |
| "learning_rate": 6.4000000000000006e-06, | |
| "loss": 0.3829, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 4.07, | |
| "learning_rate": 6.3444444444444454e-06, | |
| "loss": 0.3476, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 4.07, | |
| "learning_rate": 6.28888888888889e-06, | |
| "loss": 0.3432, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "learning_rate": 6.2333333333333335e-06, | |
| "loss": 0.3286, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "learning_rate": 6.177777777777778e-06, | |
| "loss": 0.3501, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 4.09, | |
| "learning_rate": 6.1222222222222224e-06, | |
| "loss": 0.3195, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 6.066666666666667e-06, | |
| "loss": 0.3827, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 6.011111111111112e-06, | |
| "loss": 0.8591, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 5.955555555555555e-06, | |
| "loss": 0.6715, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 5.9e-06, | |
| "loss": 0.7767, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 5.844444444444445e-06, | |
| "loss": 0.5884, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "learning_rate": 5.788888888888889e-06, | |
| "loss": 0.5195, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "learning_rate": 5.733333333333334e-06, | |
| "loss": 0.4846, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 5.677777777777779e-06, | |
| "loss": 0.4912, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 5.622222222222222e-06, | |
| "loss": 0.4949, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 5.05, | |
| "learning_rate": 5.566666666666667e-06, | |
| "loss": 0.5014, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 5.05, | |
| "learning_rate": 5.511111111111112e-06, | |
| "loss": 0.432, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 5.06, | |
| "learning_rate": 5.455555555555556e-06, | |
| "loss": 0.3757, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 5.06, | |
| "learning_rate": 5.400000000000001e-06, | |
| "loss": 0.3414, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 5.07, | |
| "learning_rate": 5.344444444444446e-06, | |
| "loss": 0.3097, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 5.07, | |
| "learning_rate": 5.288888888888889e-06, | |
| "loss": 0.3164, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 5.08, | |
| "learning_rate": 5.233333333333334e-06, | |
| "loss": 0.2891, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 5.08, | |
| "learning_rate": 5.177777777777779e-06, | |
| "loss": 0.3225, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 5.09, | |
| "learning_rate": 5.122222222222223e-06, | |
| "loss": 0.2955, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 5.0666666666666676e-06, | |
| "loss": 0.2764, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 5.011111111111111e-06, | |
| "loss": 0.8215, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 4.9555555555555565e-06, | |
| "loss": 0.6359, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 4.9000000000000005e-06, | |
| "loss": 0.7556, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 4.8444444444444446e-06, | |
| "loss": 0.607, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 4.7888888888888894e-06, | |
| "loss": 0.4868, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 4.7333333333333335e-06, | |
| "loss": 0.4199, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 6.04, | |
| "learning_rate": 4.677777777777778e-06, | |
| "loss": 0.4364, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 6.04, | |
| "learning_rate": 4.622222222222222e-06, | |
| "loss": 0.4887, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 6.05, | |
| "learning_rate": 4.566666666666667e-06, | |
| "loss": 0.4649, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 6.05, | |
| "learning_rate": 4.511111111111111e-06, | |
| "loss": 0.4186, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "learning_rate": 4.455555555555555e-06, | |
| "loss": 0.3317, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "eval_loss": 0.6079888343811035, | |
| "eval_runtime": 1490.6136, | |
| "eval_samples_per_second": 7.098, | |
| "eval_steps_per_second": 0.222, | |
| "eval_wer": 95.03539900358359, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "learning_rate": 4.4e-06, | |
| "loss": 0.3247, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 6.07, | |
| "learning_rate": 4.344444444444445e-06, | |
| "loss": 0.2809, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 6.07, | |
| "learning_rate": 4.288888888888889e-06, | |
| "loss": 0.2904, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 6.08, | |
| "learning_rate": 4.233333333333334e-06, | |
| "loss": 0.2688, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 6.08, | |
| "learning_rate": 4.177777777777778e-06, | |
| "loss": 0.2875, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 6.09, | |
| "learning_rate": 4.122222222222222e-06, | |
| "loss": 0.2781, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 6.09, | |
| "learning_rate": 4.066666666666667e-06, | |
| "loss": 0.2323, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 4.011111111111111e-06, | |
| "loss": 0.6648, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 3.955555555555556e-06, | |
| "loss": 0.7023, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 3.900000000000001e-06, | |
| "loss": 0.6855, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 3.844444444444445e-06, | |
| "loss": 0.6203, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 3.7888888888888893e-06, | |
| "loss": 0.4723, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 3.7333333333333337e-06, | |
| "loss": 0.3992, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 3.6777777777777778e-06, | |
| "loss": 0.4064, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "learning_rate": 3.6222222222222226e-06, | |
| "loss": 0.473, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "learning_rate": 3.566666666666667e-06, | |
| "loss": 0.4146, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 7.05, | |
| "learning_rate": 3.511111111111111e-06, | |
| "loss": 0.414, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 7.05, | |
| "learning_rate": 3.455555555555556e-06, | |
| "loss": 0.3208, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 7.06, | |
| "learning_rate": 3.4000000000000005e-06, | |
| "loss": 0.3054, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 7.06, | |
| "learning_rate": 3.3444444444444445e-06, | |
| "loss": 0.2648, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 7.07, | |
| "learning_rate": 3.2888888888888894e-06, | |
| "loss": 0.2672, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 7.07, | |
| "learning_rate": 3.2333333333333334e-06, | |
| "loss": 0.241, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 7.08, | |
| "learning_rate": 3.177777777777778e-06, | |
| "loss": 0.2682, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 7.08, | |
| "learning_rate": 3.1222222222222228e-06, | |
| "loss": 0.2577, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 7.09, | |
| "learning_rate": 3.066666666666667e-06, | |
| "loss": 0.2215, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 3.0111111111111113e-06, | |
| "loss": 0.5952, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 2.955555555555556e-06, | |
| "loss": 0.6769, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 2.9e-06, | |
| "loss": 0.6218, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 2.8444444444444446e-06, | |
| "loss": 0.6551, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 2.788888888888889e-06, | |
| "loss": 0.4483, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "learning_rate": 2.7333333333333336e-06, | |
| "loss": 0.3854, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "learning_rate": 2.677777777777778e-06, | |
| "loss": 0.3764, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 8.04, | |
| "learning_rate": 2.6222222222222225e-06, | |
| "loss": 0.4475, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 8.04, | |
| "learning_rate": 2.566666666666667e-06, | |
| "loss": 0.4069, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 8.05, | |
| "learning_rate": 2.5111111111111114e-06, | |
| "loss": 0.4002, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 8.05, | |
| "learning_rate": 2.455555555555556e-06, | |
| "loss": 0.3064, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 8.06, | |
| "learning_rate": 2.4000000000000003e-06, | |
| "loss": 0.2797, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 8.06, | |
| "learning_rate": 2.3444444444444448e-06, | |
| "loss": 0.2619, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 8.07, | |
| "learning_rate": 2.2888888888888892e-06, | |
| "loss": 0.2571, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 8.07, | |
| "learning_rate": 2.2333333333333333e-06, | |
| "loss": 0.225, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 8.07, | |
| "eval_loss": 0.6120802164077759, | |
| "eval_runtime": 1543.6479, | |
| "eval_samples_per_second": 6.855, | |
| "eval_steps_per_second": 0.214, | |
| "eval_wer": 91.09343588847129, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 8.08, | |
| "learning_rate": 2.1777777777777777e-06, | |
| "loss": 0.2533, | |
| "step": 4025 | |
| }, | |
| { | |
| "epoch": 8.08, | |
| "learning_rate": 2.1222222222222226e-06, | |
| "loss": 0.2407, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 8.09, | |
| "learning_rate": 2.0666666666666666e-06, | |
| "loss": 0.2141, | |
| "step": 4075 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "learning_rate": 2.011111111111111e-06, | |
| "loss": 0.5013, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 1.955555555555556e-06, | |
| "loss": 0.6818, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 1.9000000000000002e-06, | |
| "loss": 0.5907, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 9.02, | |
| "learning_rate": 1.8444444444444445e-06, | |
| "loss": 0.6743, | |
| "step": 4175 | |
| }, | |
| { | |
| "epoch": 9.02, | |
| "learning_rate": 1.788888888888889e-06, | |
| "loss": 0.4373, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 9.03, | |
| "learning_rate": 1.7333333333333336e-06, | |
| "loss": 0.3725, | |
| "step": 4225 | |
| }, | |
| { | |
| "epoch": 9.03, | |
| "learning_rate": 1.6777777777777779e-06, | |
| "loss": 0.37, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 9.04, | |
| "learning_rate": 1.6222222222222223e-06, | |
| "loss": 0.4152, | |
| "step": 4275 | |
| }, | |
| { | |
| "epoch": 9.04, | |
| "learning_rate": 1.566666666666667e-06, | |
| "loss": 0.396, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 9.05, | |
| "learning_rate": 1.5111111111111112e-06, | |
| "loss": 0.3896, | |
| "step": 4325 | |
| }, | |
| { | |
| "epoch": 9.05, | |
| "learning_rate": 1.4555555555555557e-06, | |
| "loss": 0.3001, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 9.06, | |
| "learning_rate": 1.4000000000000001e-06, | |
| "loss": 0.2766, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 9.06, | |
| "learning_rate": 1.3444444444444446e-06, | |
| "loss": 0.2556, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 9.07, | |
| "learning_rate": 1.288888888888889e-06, | |
| "loss": 0.2457, | |
| "step": 4425 | |
| }, | |
| { | |
| "epoch": 9.07, | |
| "learning_rate": 1.2333333333333335e-06, | |
| "loss": 0.2149, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 9.08, | |
| "learning_rate": 1.1777777777777778e-06, | |
| "loss": 0.2275, | |
| "step": 4475 | |
| }, | |
| { | |
| "epoch": 9.08, | |
| "learning_rate": 1.1222222222222222e-06, | |
| "loss": 0.2381, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 9.09, | |
| "learning_rate": 1.066666666666667e-06, | |
| "loss": 0.2145, | |
| "step": 4525 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "learning_rate": 1.0111111111111111e-06, | |
| "loss": 0.3767, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 10.01, | |
| "learning_rate": 9.555555555555556e-07, | |
| "loss": 0.7333, | |
| "step": 4575 | |
| }, | |
| { | |
| "epoch": 10.01, | |
| "learning_rate": 9.000000000000001e-07, | |
| "loss": 0.5436, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 10.02, | |
| "learning_rate": 8.444444444444445e-07, | |
| "loss": 0.6899, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 10.02, | |
| "learning_rate": 7.888888888888889e-07, | |
| "loss": 0.4377, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 10.03, | |
| "learning_rate": 7.333333333333334e-07, | |
| "loss": 0.368, | |
| "step": 4675 | |
| }, | |
| { | |
| "epoch": 10.03, | |
| "learning_rate": 6.777777777777779e-07, | |
| "loss": 0.376, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 10.04, | |
| "learning_rate": 6.222222222222223e-07, | |
| "loss": 0.4094, | |
| "step": 4725 | |
| }, | |
| { | |
| "epoch": 10.04, | |
| "learning_rate": 5.666666666666667e-07, | |
| "loss": 0.3762, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 10.05, | |
| "learning_rate": 5.111111111111112e-07, | |
| "loss": 0.3765, | |
| "step": 4775 | |
| }, | |
| { | |
| "epoch": 10.05, | |
| "learning_rate": 4.5555555555555563e-07, | |
| "loss": 0.3095, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 10.06, | |
| "learning_rate": 4.0000000000000003e-07, | |
| "loss": 0.2658, | |
| "step": 4825 | |
| }, | |
| { | |
| "epoch": 10.06, | |
| "learning_rate": 3.444444444444445e-07, | |
| "loss": 0.257, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 10.07, | |
| "learning_rate": 2.888888888888889e-07, | |
| "loss": 0.2347, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 10.07, | |
| "learning_rate": 2.3333333333333336e-07, | |
| "loss": 0.2189, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 10.08, | |
| "learning_rate": 1.777777777777778e-07, | |
| "loss": 0.2174, | |
| "step": 4925 | |
| }, | |
| { | |
| "epoch": 10.08, | |
| "learning_rate": 1.2222222222222225e-07, | |
| "loss": 0.2355, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 10.09, | |
| "learning_rate": 6.666666666666668e-08, | |
| "loss": 0.208, | |
| "step": 4975 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "learning_rate": 1.1111111111111112e-08, | |
| "loss": 0.3166, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_loss": 0.6091899275779724, | |
| "eval_runtime": 1613.9377, | |
| "eval_samples_per_second": 6.556, | |
| "eval_steps_per_second": 0.205, | |
| "eval_wer": 92.31710514815138, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "step": 5000, | |
| "total_flos": 7.8780432384e+18, | |
| "train_loss": 0.5324803882598877, | |
| "train_runtime": 32150.3272, | |
| "train_samples_per_second": 9.953, | |
| "train_steps_per_second": 0.156 | |
| } | |
| ], | |
| "max_steps": 5000, | |
| "num_train_epochs": 9223372036854775807, | |
| "total_flos": 7.8780432384e+18, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |