| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 1.0, | |
| "global_step": 50, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 5e-08, | |
| "loss": 1.3906, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.347826086956522e-08, | |
| "loss": 1.3831, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 8.26086956521739e-08, | |
| "loss": 1.406, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 7.173913043478261e-08, | |
| "loss": 1.4041, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 6.086956521739131e-08, | |
| "loss": 1.3447, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 5e-08, | |
| "loss": 1.3263, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 3.913043478260869e-08, | |
| "loss": 1.3201, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.8260869565217388e-08, | |
| "loss": 1.3223, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.7391304347826087e-08, | |
| "loss": 1.2421, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 6.521739130434782e-09, | |
| "loss": 1.3366, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "step": 50, | |
| "total_flos": 7.8780432384e+16, | |
| "train_loss": 1.3475898265838624, | |
| "train_runtime": 419.0595, | |
| "train_samples_per_second": 7.636, | |
| "train_steps_per_second": 0.119 | |
| } | |
| ], | |
| "max_steps": 50, | |
| "num_train_epochs": 9223372036854775807, | |
| "total_flos": 7.8780432384e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |