| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.00024246644326596082, | |
| "eval_steps": 13, | |
| "global_step": 39, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 6.2170882888707906e-06, | |
| "eval_loss": NaN, | |
| "eval_runtime": 1431.5295, | |
| "eval_samples_per_second": 23.655, | |
| "eval_steps_per_second": 11.828, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 6.21708828887079e-05, | |
| "grad_norm": NaN, | |
| "learning_rate": 0.0002, | |
| "loss": 1.7859, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 8.082214775532027e-05, | |
| "eval_loss": NaN, | |
| "eval_runtime": 1429.1438, | |
| "eval_samples_per_second": 23.695, | |
| "eval_steps_per_second": 11.848, | |
| "step": 13 | |
| }, | |
| { | |
| "epoch": 0.0001243417657774158, | |
| "grad_norm": NaN, | |
| "learning_rate": 0.0002, | |
| "loss": 0.3708, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.00016164429551064055, | |
| "eval_loss": NaN, | |
| "eval_runtime": 1430.9366, | |
| "eval_samples_per_second": 23.665, | |
| "eval_steps_per_second": 11.833, | |
| "step": 26 | |
| }, | |
| { | |
| "epoch": 0.0001865126486661237, | |
| "grad_norm": NaN, | |
| "learning_rate": 0.0002, | |
| "loss": 0.133, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.00024246644326596082, | |
| "eval_loss": NaN, | |
| "eval_runtime": 1429.335, | |
| "eval_samples_per_second": 23.691, | |
| "eval_steps_per_second": 11.846, | |
| "step": 39 | |
| } | |
| ], | |
| "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": false | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 4467957002403840.0, | |
| "train_batch_size": 2, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |