| { | |
| "best_metric": 3.9855642318725586, | |
| "best_model_checkpoint": "output/epic-rap-battles-of-history/checkpoint-50", | |
| "epoch": 2.0, | |
| "global_step": 50, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0001240985658141214, | |
| "loss": 4.5824, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 8.97985658141214e-05, | |
| "loss": 4.2385, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 4.740143418587861e-05, | |
| "loss": 4.1226, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.3101434185878613e-05, | |
| "loss": 4.2038, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 0.0, | |
| "loss": 4.1695, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_loss": 4.043574810028076, | |
| "eval_runtime": 1.45, | |
| "eval_samples_per_second": 22.759, | |
| "eval_steps_per_second": 3.448, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 1.3101434185878598e-05, | |
| "loss": 3.9483, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 4.74014341858786e-05, | |
| "loss": 4.0002, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 8.979856581412138e-05, | |
| "loss": 3.8185, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 0.0001240985658141214, | |
| "loss": 3.9126, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 0.0001372, | |
| "loss": 3.8351, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_loss": 3.9855642318725586, | |
| "eval_runtime": 1.4748, | |
| "eval_samples_per_second": 22.375, | |
| "eval_steps_per_second": 3.39, | |
| "step": 50 | |
| } | |
| ], | |
| "max_steps": 625, | |
| "num_train_epochs": 25, | |
| "total_flos": 51735822336000.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |