| { | |
| "best_global_step": null, | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 4.12796697626419, | |
| "eval_steps": 1000, | |
| "global_step": 4000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0319917440660475, | |
| "grad_norm": 0.6404266953468323, | |
| "learning_rate": 5.154798761609907e-05, | |
| "loss": 0.3551, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.0319917440660475, | |
| "eval_loss": 0.2621232867240906, | |
| "eval_runtime": 285.996, | |
| "eval_samples_per_second": 6.025, | |
| "eval_steps_per_second": 0.755, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.063983488132095, | |
| "grad_norm": 0.5686078071594238, | |
| "learning_rate": 9.9996982116359e-05, | |
| "loss": 0.2362, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 2.063983488132095, | |
| "eval_loss": 0.2504255175590515, | |
| "eval_runtime": 285.5907, | |
| "eval_samples_per_second": 6.033, | |
| "eval_steps_per_second": 0.756, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 3.0959752321981426, | |
| "grad_norm": 0.510012686252594, | |
| "learning_rate": 9.908976018041207e-05, | |
| "loss": 0.2221, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 3.0959752321981426, | |
| "eval_loss": 0.2474418580532074, | |
| "eval_runtime": 285.7076, | |
| "eval_samples_per_second": 6.031, | |
| "eval_steps_per_second": 0.756, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 4.12796697626419, | |
| "grad_norm": 0.4461008608341217, | |
| "learning_rate": 9.659427070509466e-05, | |
| "loss": 0.2102, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 4.12796697626419, | |
| "eval_loss": 0.247001051902771, | |
| "eval_runtime": 285.5088, | |
| "eval_samples_per_second": 6.035, | |
| "eval_steps_per_second": 0.757, | |
| "step": 4000 | |
| } | |
| ], | |
| "logging_steps": 1000, | |
| "max_steps": 19380, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 20, | |
| "save_steps": 1000, | |
| "stateful_callbacks": { | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": false | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 1.8354072285918265e+19, | |
| "train_batch_size": 1, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |