| { | |
| "best_metric": 15.541324622856406, | |
| "best_model_checkpoint": "/scratch/project_465000370/asr_children/models/whisper-large-et-clinic/checkpoint-2000", | |
| "epoch": 16.12806403201601, | |
| "global_step": 2000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.1500000000000002e-06, | |
| "loss": 0.495, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 2.4000000000000003e-06, | |
| "loss": 0.2703, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.65e-06, | |
| "loss": 0.2068, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.9000000000000005e-06, | |
| "loss": 0.2183, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 6.15e-06, | |
| "loss": 0.1988, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 7.4e-06, | |
| "loss": 0.1218, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 8.65e-06, | |
| "loss": 0.1199, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 9.9e-06, | |
| "loss": 0.1196, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 9.872222222222223e-06, | |
| "loss": 0.1177, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 9.733333333333334e-06, | |
| "loss": 0.1277, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 9.594444444444445e-06, | |
| "loss": 0.0481, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 9.455555555555557e-06, | |
| "loss": 0.0542, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 9.316666666666667e-06, | |
| "loss": 0.06, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 9.17777777777778e-06, | |
| "loss": 0.0543, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 9.038888888888889e-06, | |
| "loss": 0.0481, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 8.900000000000001e-06, | |
| "loss": 0.0238, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 8.761111111111112e-06, | |
| "loss": 0.0236, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 8.627777777777778e-06, | |
| "loss": 0.0265, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 8.48888888888889e-06, | |
| "loss": 0.0277, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 8.35e-06, | |
| "loss": 0.0278, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "eval_loss": 0.2985570430755615, | |
| "eval_runtime": 2329.8998, | |
| "eval_samples_per_second": 0.552, | |
| "eval_steps_per_second": 0.552, | |
| "eval_wer": 16.99402587355482, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 4.23, | |
| "learning_rate": 8.211111111111112e-06, | |
| "loss": 0.013, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 4.43, | |
| "learning_rate": 8.072222222222223e-06, | |
| "loss": 0.0138, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 4.63, | |
| "learning_rate": 7.933333333333334e-06, | |
| "loss": 0.0165, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 4.83, | |
| "learning_rate": 7.794444444444445e-06, | |
| "loss": 0.0145, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 7.655555555555556e-06, | |
| "loss": 0.0143, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 5.24, | |
| "learning_rate": 7.516666666666668e-06, | |
| "loss": 0.008, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 5.44, | |
| "learning_rate": 7.377777777777778e-06, | |
| "loss": 0.0103, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 5.64, | |
| "learning_rate": 7.23888888888889e-06, | |
| "loss": 0.0081, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 5.84, | |
| "learning_rate": 7.100000000000001e-06, | |
| "loss": 0.0083, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 6.05, | |
| "learning_rate": 6.9611111111111116e-06, | |
| "loss": 0.0087, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 6.25, | |
| "learning_rate": 6.8222222222222225e-06, | |
| "loss": 0.007, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 6.45, | |
| "learning_rate": 6.683333333333334e-06, | |
| "loss": 0.0054, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 6.65, | |
| "learning_rate": 6.544444444444445e-06, | |
| "loss": 0.0072, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 6.85, | |
| "learning_rate": 6.405555555555556e-06, | |
| "loss": 0.0046, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 7.06, | |
| "learning_rate": 6.266666666666668e-06, | |
| "loss": 0.0053, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 7.26, | |
| "learning_rate": 6.127777777777778e-06, | |
| "loss": 0.0036, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 7.46, | |
| "learning_rate": 5.98888888888889e-06, | |
| "loss": 0.0035, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 7.66, | |
| "learning_rate": 5.85e-06, | |
| "loss": 0.0039, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 7.86, | |
| "learning_rate": 5.711111111111112e-06, | |
| "loss": 0.004, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 8.06, | |
| "learning_rate": 5.572222222222223e-06, | |
| "loss": 0.0037, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 8.06, | |
| "eval_loss": 0.3382585346698761, | |
| "eval_runtime": 2322.4054, | |
| "eval_samples_per_second": 0.554, | |
| "eval_steps_per_second": 0.554, | |
| "eval_wer": 16.17741866162376, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 8.26, | |
| "learning_rate": 5.4333333333333335e-06, | |
| "loss": 0.002, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 8.46, | |
| "learning_rate": 5.294444444444445e-06, | |
| "loss": 0.0036, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 8.66, | |
| "learning_rate": 5.155555555555556e-06, | |
| "loss": 0.0029, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 8.86, | |
| "learning_rate": 5.016666666666667e-06, | |
| "loss": 0.0026, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 9.07, | |
| "learning_rate": 4.877777777777778e-06, | |
| "loss": 0.0023, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 9.27, | |
| "learning_rate": 4.73888888888889e-06, | |
| "loss": 0.0023, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 9.47, | |
| "learning_rate": 4.600000000000001e-06, | |
| "loss": 0.0023, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 9.67, | |
| "learning_rate": 4.461111111111112e-06, | |
| "loss": 0.0019, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 9.87, | |
| "learning_rate": 4.322222222222223e-06, | |
| "loss": 0.002, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 10.08, | |
| "learning_rate": 4.183333333333334e-06, | |
| "loss": 0.0017, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 10.28, | |
| "learning_rate": 4.044444444444445e-06, | |
| "loss": 0.0014, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 10.48, | |
| "learning_rate": 3.9055555555555555e-06, | |
| "loss": 0.0021, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 10.68, | |
| "learning_rate": 3.766666666666667e-06, | |
| "loss": 0.0017, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 10.88, | |
| "learning_rate": 3.627777777777778e-06, | |
| "loss": 0.0024, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 11.09, | |
| "learning_rate": 3.4888888888888896e-06, | |
| "loss": 0.0015, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 11.29, | |
| "learning_rate": 3.3500000000000005e-06, | |
| "loss": 0.0012, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 11.49, | |
| "learning_rate": 3.2111111111111115e-06, | |
| "loss": 0.0012, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 11.69, | |
| "learning_rate": 3.0722222222222224e-06, | |
| "loss": 0.0017, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 11.89, | |
| "learning_rate": 2.9333333333333338e-06, | |
| "loss": 0.0022, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 12.1, | |
| "learning_rate": 2.7944444444444447e-06, | |
| "loss": 0.0017, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 12.1, | |
| "eval_loss": 0.3618781566619873, | |
| "eval_runtime": 2327.1444, | |
| "eval_samples_per_second": 0.553, | |
| "eval_steps_per_second": 0.553, | |
| "eval_wer": 15.803498517213221, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 12.3, | |
| "learning_rate": 2.6555555555555556e-06, | |
| "loss": 0.0012, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 12.5, | |
| "learning_rate": 2.5166666666666666e-06, | |
| "loss": 0.0016, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 12.7, | |
| "learning_rate": 2.377777777777778e-06, | |
| "loss": 0.0016, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 12.9, | |
| "learning_rate": 2.238888888888889e-06, | |
| "loss": 0.0012, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 13.1, | |
| "learning_rate": 2.1000000000000002e-06, | |
| "loss": 0.0012, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 13.3, | |
| "learning_rate": 1.9611111111111116e-06, | |
| "loss": 0.0011, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 13.5, | |
| "learning_rate": 1.8222222222222225e-06, | |
| "loss": 0.0014, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 13.7, | |
| "learning_rate": 1.6833333333333335e-06, | |
| "loss": 0.0012, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 13.9, | |
| "learning_rate": 1.5444444444444446e-06, | |
| "loss": 0.0013, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 14.11, | |
| "learning_rate": 1.4055555555555555e-06, | |
| "loss": 0.0013, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 14.31, | |
| "learning_rate": 1.2666666666666669e-06, | |
| "loss": 0.0012, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 14.51, | |
| "learning_rate": 1.1277777777777778e-06, | |
| "loss": 0.001, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 14.71, | |
| "learning_rate": 9.88888888888889e-07, | |
| "loss": 0.001, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 14.91, | |
| "learning_rate": 8.500000000000001e-07, | |
| "loss": 0.0011, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 15.12, | |
| "learning_rate": 7.111111111111112e-07, | |
| "loss": 0.0011, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 15.32, | |
| "learning_rate": 5.722222222222223e-07, | |
| "loss": 0.001, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 15.52, | |
| "learning_rate": 4.333333333333334e-07, | |
| "loss": 0.0009, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 15.72, | |
| "learning_rate": 2.9444444444444444e-07, | |
| "loss": 0.0012, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 15.92, | |
| "learning_rate": 1.5555555555555556e-07, | |
| "loss": 0.0012, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 16.13, | |
| "learning_rate": 1.666666666666667e-08, | |
| "loss": 0.0008, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 16.13, | |
| "eval_loss": 0.3861992359161377, | |
| "eval_runtime": 2318.6816, | |
| "eval_samples_per_second": 0.555, | |
| "eval_steps_per_second": 0.555, | |
| "eval_wer": 15.541324622856406, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 16.13, | |
| "step": 2000, | |
| "total_flos": 1.36902067421184e+20, | |
| "train_loss": 0.032722438401542604, | |
| "train_runtime": 36967.2072, | |
| "train_samples_per_second": 1.731, | |
| "train_steps_per_second": 0.054 | |
| } | |
| ], | |
| "max_steps": 2000, | |
| "num_train_epochs": 17, | |
| "total_flos": 1.36902067421184e+20, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |