| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.8394543546694648, | |
| "eval_steps": 10, | |
| "global_step": 50, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.016789087093389297, | |
| "eval_loss": NaN, | |
| "eval_runtime": 13.6492, | |
| "eval_samples_per_second": 29.452, | |
| "eval_steps_per_second": 3.736, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.16789087093389296, | |
| "grad_norm": NaN, | |
| "learning_rate": 9.330127018922194e-05, | |
| "loss": 0.0, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.16789087093389296, | |
| "eval_loss": NaN, | |
| "eval_runtime": 12.6712, | |
| "eval_samples_per_second": 31.725, | |
| "eval_steps_per_second": 4.025, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.3357817418677859, | |
| "grad_norm": NaN, | |
| "learning_rate": 6.91341716182545e-05, | |
| "loss": 0.0, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.3357817418677859, | |
| "eval_loss": NaN, | |
| "eval_runtime": 12.5667, | |
| "eval_samples_per_second": 31.989, | |
| "eval_steps_per_second": 4.058, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.5036726128016789, | |
| "grad_norm": NaN, | |
| "learning_rate": 3.705904774487396e-05, | |
| "loss": 0.0, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.5036726128016789, | |
| "eval_loss": NaN, | |
| "eval_runtime": 12.6016, | |
| "eval_samples_per_second": 31.901, | |
| "eval_steps_per_second": 4.047, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.6715634837355718, | |
| "grad_norm": NaN, | |
| "learning_rate": 1.0332332985438248e-05, | |
| "loss": 0.0, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.6715634837355718, | |
| "eval_loss": NaN, | |
| "eval_runtime": 12.5994, | |
| "eval_samples_per_second": 31.906, | |
| "eval_steps_per_second": 4.048, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.8394543546694648, | |
| "grad_norm": NaN, | |
| "learning_rate": 0.0, | |
| "loss": 0.0, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.8394543546694648, | |
| "eval_loss": NaN, | |
| "eval_runtime": 12.6624, | |
| "eval_samples_per_second": 31.748, | |
| "eval_steps_per_second": 4.028, | |
| "step": 50 | |
| } | |
| ], | |
| "logging_steps": 10, | |
| "max_steps": 50, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 1, | |
| "save_steps": 13, | |
| "stateful_callbacks": { | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": true | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 5.48485308874752e+16, | |
| "train_batch_size": 8, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |