| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 999.001, | |
| "global_step": 1000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 24.0, | |
| "learning_rate": 6.808639180087963e-06, | |
| "loss": 0.2617, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 49.0, | |
| "learning_rate": 8.406206186877936e-06, | |
| "loss": 0.0001, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 74.0, | |
| "learning_rate": 9.316614300602277e-06, | |
| "loss": 0.0, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 99.0, | |
| "learning_rate": 9.956130378462474e-06, | |
| "loss": 0.0, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 124.0, | |
| "learning_rate": 9.755555555555556e-06, | |
| "loss": 0.0, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 149.0, | |
| "learning_rate": 9.47777777777778e-06, | |
| "loss": 0.0, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 174.0, | |
| "learning_rate": 9.200000000000002e-06, | |
| "loss": 0.0, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 199.0, | |
| "learning_rate": 8.922222222222224e-06, | |
| "loss": 0.0, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 224.0, | |
| "learning_rate": 8.644444444444445e-06, | |
| "loss": 0.0, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 249.0, | |
| "learning_rate": 8.366666666666667e-06, | |
| "loss": 0.0, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 274.0, | |
| "learning_rate": 8.08888888888889e-06, | |
| "loss": 0.0, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 299.0, | |
| "learning_rate": 7.811111111111111e-06, | |
| "loss": 0.0, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 324.0, | |
| "learning_rate": 7.533333333333334e-06, | |
| "loss": 0.0, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 349.0, | |
| "learning_rate": 7.255555555555556e-06, | |
| "loss": 0.0, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 374.0, | |
| "learning_rate": 6.977777777777779e-06, | |
| "loss": 0.0, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 399.0, | |
| "learning_rate": 6.700000000000001e-06, | |
| "loss": 0.0, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 424.0, | |
| "learning_rate": 6.422222222222223e-06, | |
| "loss": 0.0, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 449.0, | |
| "learning_rate": 6.144444444444445e-06, | |
| "loss": 0.0, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 474.0, | |
| "learning_rate": 5.8666666666666675e-06, | |
| "loss": 0.0, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 499.0, | |
| "learning_rate": 5.588888888888889e-06, | |
| "loss": 0.0, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 524.0, | |
| "learning_rate": 5.311111111111111e-06, | |
| "loss": 0.0, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 549.0, | |
| "learning_rate": 5.033333333333333e-06, | |
| "loss": 0.0, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 574.0, | |
| "learning_rate": 4.755555555555556e-06, | |
| "loss": 0.0, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 599.0, | |
| "learning_rate": 4.477777777777778e-06, | |
| "loss": 0.0, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 624.0, | |
| "learning_rate": 4.2000000000000004e-06, | |
| "loss": 0.0, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 649.0, | |
| "learning_rate": 3.922222222222223e-06, | |
| "loss": 0.0, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 674.0, | |
| "learning_rate": 3.644444444444445e-06, | |
| "loss": 0.0, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 699.0, | |
| "learning_rate": 3.366666666666667e-06, | |
| "loss": 0.0, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 724.0, | |
| "learning_rate": 3.088888888888889e-06, | |
| "loss": 0.0, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 749.0, | |
| "learning_rate": 2.811111111111111e-06, | |
| "loss": 0.0, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 774.0, | |
| "learning_rate": 2.5333333333333338e-06, | |
| "loss": 0.0, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 799.0, | |
| "learning_rate": 2.2555555555555557e-06, | |
| "loss": 0.0, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 824.0, | |
| "learning_rate": 1.977777777777778e-06, | |
| "loss": 0.0, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 849.0, | |
| "learning_rate": 1.7000000000000002e-06, | |
| "loss": 0.0, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 874.0, | |
| "learning_rate": 1.4222222222222223e-06, | |
| "loss": 0.0, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 899.0, | |
| "learning_rate": 1.1444444444444446e-06, | |
| "loss": 0.0, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 924.0, | |
| "learning_rate": 8.666666666666668e-07, | |
| "loss": 0.0, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 949.0, | |
| "learning_rate": 5.888888888888889e-07, | |
| "loss": 0.0, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 974.0, | |
| "learning_rate": 3.111111111111111e-07, | |
| "loss": 0.0, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 999.0, | |
| "learning_rate": 3.333333333333334e-08, | |
| "loss": 0.0, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 999.0, | |
| "eval_loss": 1.201171875, | |
| "eval_runtime": 4.5518, | |
| "eval_samples_per_second": 0.439, | |
| "eval_steps_per_second": 0.22, | |
| "eval_wer": 14.634146341463413, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 999.0, | |
| "step": 1000, | |
| "total_flos": 1.4862197628694692e+19, | |
| "train_loss": 0.006550703465938568, | |
| "train_runtime": 3578.7906, | |
| "train_samples_per_second": 8.942, | |
| "train_steps_per_second": 0.279 | |
| } | |
| ], | |
| "max_steps": 1000, | |
| "num_train_epochs": 9223372036854775807, | |
| "total_flos": 1.4862197628694692e+19, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |