| { |
| "best_metric": 0.9423573608828364, |
| "best_model_checkpoint": "model_saves/deberta-v3-large_lemon-spell_10k_1_p3/checkpoint-268", |
| "epoch": 4.0, |
| "global_step": 1072, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 1.0, |
| "eval_accuracy": 0.9423573608828364, |
| "eval_loss": 0.42043185234069824, |
| "eval_runtime": 4.5835, |
| "eval_samples_per_second": 956.038, |
| "eval_steps_per_second": 7.636, |
| "step": 268 |
| }, |
| { |
| "epoch": 2.0, |
| "eval_accuracy": 0.9420666599581838, |
| "eval_loss": 0.4363779127597809, |
| "eval_runtime": 4.5594, |
| "eval_samples_per_second": 961.093, |
| "eval_steps_per_second": 7.676, |
| "step": 536 |
| }, |
| { |
| "epoch": 3.0, |
| "eval_accuracy": 0.9417256454119568, |
| "eval_loss": 0.46696099638938904, |
| "eval_runtime": 4.5609, |
| "eval_samples_per_second": 960.769, |
| "eval_steps_per_second": 7.674, |
| "step": 804 |
| }, |
| { |
| "epoch": 3.73, |
| "learning_rate": 1e-05, |
| "loss": 0.2897, |
| "step": 1000 |
| }, |
| { |
| "epoch": 4.0, |
| "eval_accuracy": 0.9405572513109494, |
| "eval_loss": 0.4973376393318176, |
| "eval_runtime": 4.5634, |
| "eval_samples_per_second": 960.256, |
| "eval_steps_per_second": 7.67, |
| "step": 1072 |
| }, |
| { |
| "epoch": 4.0, |
| "step": 1072, |
| "total_flos": 1.3785700008722432e+16, |
| "train_loss": 0.28410458475796146, |
| "train_runtime": 490.8817, |
| "train_samples_per_second": 1048.236, |
| "train_steps_per_second": 8.189 |
| } |
| ], |
| "max_steps": 4020, |
| "num_train_epochs": 15, |
| "total_flos": 1.3785700008722432e+16, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
