| { |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 6.0, |
| "global_step": 1650, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 1.0, |
| "learning_rate": 3e-05, |
| "loss": 1.8366, |
| "step": 275 |
| }, |
| { |
| "epoch": 1.0, |
| "eval_exact_match": 53.06051066806576, |
| "eval_f1": 61.177359074813886, |
| "eval_runtime": 3.5321, |
| "eval_samples_per_second": 2428.33, |
| "eval_steps_per_second": 7.644, |
| "step": 275 |
| }, |
| { |
| "epoch": 2.0, |
| "learning_rate": 3e-05, |
| "loss": 1.0078, |
| "step": 550 |
| }, |
| { |
| "epoch": 2.0, |
| "eval_exact_match": 54.58785123003381, |
| "eval_f1": 62.44318229159688, |
| "eval_runtime": 2.9617, |
| "eval_samples_per_second": 2895.929, |
| "eval_steps_per_second": 9.116, |
| "step": 550 |
| }, |
| { |
| "epoch": 3.0, |
| "learning_rate": 3e-05, |
| "loss": 0.8045, |
| "step": 825 |
| }, |
| { |
| "epoch": 3.0, |
| "eval_exact_match": 54.40130581788504, |
| "eval_f1": 62.934647790726544, |
| "eval_runtime": 3.0061, |
| "eval_samples_per_second": 2853.196, |
| "eval_steps_per_second": 8.982, |
| "step": 825 |
| }, |
| { |
| "epoch": 4.0, |
| "learning_rate": 3e-05, |
| "loss": 0.6601, |
| "step": 1100 |
| }, |
| { |
| "epoch": 4.0, |
| "eval_exact_match": 54.821032995219774, |
| "eval_f1": 62.74065134891623, |
| "eval_runtime": 2.9452, |
| "eval_samples_per_second": 2912.204, |
| "eval_steps_per_second": 9.167, |
| "step": 1100 |
| }, |
| { |
| "epoch": 5.0, |
| "learning_rate": 3e-05, |
| "loss": 0.5408, |
| "step": 1375 |
| }, |
| { |
| "epoch": 5.0, |
| "eval_exact_match": 54.72776028914539, |
| "eval_f1": 62.348767485436014, |
| "eval_runtime": 2.9711, |
| "eval_samples_per_second": 2886.821, |
| "eval_steps_per_second": 9.088, |
| "step": 1375 |
| }, |
| { |
| "epoch": 6.0, |
| "learning_rate": 3e-05, |
| "loss": 0.448, |
| "step": 1650 |
| }, |
| { |
| "epoch": 6.0, |
| "eval_exact_match": 56.24344176285415, |
| "eval_f1": 63.68426396891711, |
| "eval_runtime": 3.087, |
| "eval_samples_per_second": 2778.408, |
| "eval_steps_per_second": 8.746, |
| "step": 1650 |
| } |
| ], |
| "max_steps": 2750, |
| "num_train_epochs": 10, |
| "total_flos": 198652180037632.0, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
