| { |
| "best_metric": 0.9400820671071903, |
| "best_model_checkpoint": "model_saves/electra-large_lemon-spell_10k_2_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": 0.9400820671071903, |
| "eval_loss": 0.4438272714614868, |
| "eval_runtime": 3.4736, |
| "eval_samples_per_second": 1261.504, |
| "eval_steps_per_second": 10.076, |
| "step": 268 |
| }, |
| { |
| "epoch": 2.0, |
| "eval_accuracy": 0.9392267355404242, |
| "eval_loss": 0.4589129388332367, |
| "eval_runtime": 3.4779, |
| "eval_samples_per_second": 1259.94, |
| "eval_steps_per_second": 10.063, |
| "step": 536 |
| }, |
| { |
| "epoch": 3.0, |
| "eval_accuracy": 0.9391316986996724, |
| "eval_loss": 0.4756287932395935, |
| "eval_runtime": 3.4931, |
| "eval_samples_per_second": 1254.456, |
| "eval_steps_per_second": 10.02, |
| "step": 804 |
| }, |
| { |
| "epoch": 3.73, |
| "learning_rate": 1e-05, |
| "loss": 0.3213, |
| "step": 1000 |
| }, |
| { |
| "epoch": 4.0, |
| "eval_accuracy": 0.9390757946757008, |
| "eval_loss": 0.5057807564735413, |
| "eval_runtime": 3.473, |
| "eval_samples_per_second": 1261.734, |
| "eval_steps_per_second": 10.078, |
| "step": 1072 |
| }, |
| { |
| "epoch": 4.0, |
| "step": 1072, |
| "total_flos": 1.4147605630025728e+16, |
| "train_loss": 0.31617475445590804, |
| "train_runtime": 319.6414, |
| "train_samples_per_second": 1609.804, |
| "train_steps_per_second": 12.577 |
| } |
| ], |
| "max_steps": 4020, |
| "num_train_epochs": 15, |
| "total_flos": 1.4147605630025728e+16, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
