| { | |
| "best_metric": 1.0, | |
| "best_model_checkpoint": "finetuned\\checkpoint-5", | |
| "epoch": 10.0, | |
| "eval_steps": 500, | |
| "global_step": 50, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 5e-05, | |
| "loss": 0.7055, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 1.0, | |
| "eval_loss": 0.6571568846702576, | |
| "eval_runtime": 0.0813, | |
| "eval_samples_per_second": 24.587, | |
| "eval_steps_per_second": 12.294, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 4.4444444444444447e-05, | |
| "loss": 0.667, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.0, | |
| "eval_loss": 0.9093643426895142, | |
| "eval_runtime": 0.0391, | |
| "eval_samples_per_second": 51.099, | |
| "eval_steps_per_second": 25.549, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 3.888888888888889e-05, | |
| "loss": 0.6129, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 1.0, | |
| "eval_loss": 0.6191004514694214, | |
| "eval_runtime": 0.0397, | |
| "eval_samples_per_second": 50.333, | |
| "eval_steps_per_second": 25.167, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 3.3333333333333335e-05, | |
| "loss": 0.6345, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.5, | |
| "eval_loss": 0.6391083002090454, | |
| "eval_runtime": 0.0399, | |
| "eval_samples_per_second": 50.142, | |
| "eval_steps_per_second": 25.071, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 2.777777777777778e-05, | |
| "loss": 0.5029, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.5, | |
| "eval_loss": 0.6429222822189331, | |
| "eval_runtime": 0.0473, | |
| "eval_samples_per_second": 42.326, | |
| "eval_steps_per_second": 21.163, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 2.2222222222222223e-05, | |
| "loss": 0.4698, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.5, | |
| "eval_loss": 0.6175457239151001, | |
| "eval_runtime": 0.0518, | |
| "eval_samples_per_second": 38.587, | |
| "eval_steps_per_second": 19.294, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 1.6666666666666667e-05, | |
| "loss": 0.3119, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 1.0, | |
| "eval_loss": 0.43689876794815063, | |
| "eval_runtime": 0.041, | |
| "eval_samples_per_second": 48.823, | |
| "eval_steps_per_second": 24.411, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 1.1111111111111112e-05, | |
| "loss": 0.286, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.5, | |
| "eval_loss": 0.7379075884819031, | |
| "eval_runtime": 0.0397, | |
| "eval_samples_per_second": 50.369, | |
| "eval_steps_per_second": 25.185, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "learning_rate": 5.555555555555556e-06, | |
| "loss": 0.2241, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.5, | |
| "eval_loss": 0.8524875044822693, | |
| "eval_runtime": 0.0431, | |
| "eval_samples_per_second": 46.432, | |
| "eval_steps_per_second": 23.216, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "learning_rate": 0.0, | |
| "loss": 0.1981, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_accuracy": 0.5, | |
| "eval_loss": 0.8950199484825134, | |
| "eval_runtime": 0.0399, | |
| "eval_samples_per_second": 50.108, | |
| "eval_steps_per_second": 25.054, | |
| "step": 50 | |
| } | |
| ], | |
| "logging_steps": 5, | |
| "max_steps": 50, | |
| "num_train_epochs": 10, | |
| "save_steps": 500, | |
| "total_flos": 1539644174588520.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |