| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.9748964172556666, | |
| "eval_steps": 500, | |
| "global_step": 4000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 4.879469122426869e-05, | |
| "loss": 0.3842, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "eval_accuracy": 0.9771378667109926, | |
| "eval_f1": 0.8381009137862535, | |
| "eval_loss": 0.07010635733604431, | |
| "eval_precision": 0.8681427219227129, | |
| "eval_recall": 0.8100687377596867, | |
| "eval_runtime": 73.0194, | |
| "eval_samples_per_second": 224.735, | |
| "eval_steps_per_second": 7.026, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 4.202329360780065e-05, | |
| "loss": 0.0597, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "eval_accuracy": 0.9804913313861603, | |
| "eval_f1": 0.8726274509803922, | |
| "eval_loss": 0.05635998398065567, | |
| "eval_precision": 0.8915421290917104, | |
| "eval_recall": 0.8544986751660842, | |
| "eval_runtime": 75.0428, | |
| "eval_samples_per_second": 218.675, | |
| "eval_steps_per_second": 6.836, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 3.525189599133261e-05, | |
| "loss": 0.0524, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "eval_accuracy": 0.9802354021387242, | |
| "eval_f1": 0.8719290536491832, | |
| "eval_loss": 0.05560198798775673, | |
| "eval_precision": 0.9085119667013527, | |
| "eval_recall": 0.8381782573633885, | |
| "eval_runtime": 72.6323, | |
| "eval_samples_per_second": 225.933, | |
| "eval_steps_per_second": 7.063, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.8480498374864572e-05, | |
| "loss": 0.0448, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "eval_accuracy": 0.9826548569830427, | |
| "eval_f1": 0.8893414799728445, | |
| "eval_loss": 0.04828519746661186, | |
| "eval_precision": 0.8985262992866662, | |
| "eval_recall": 0.8803425367689413, | |
| "eval_runtime": 74.6781, | |
| "eval_samples_per_second": 219.743, | |
| "eval_steps_per_second": 6.869, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 2.1709100758396535e-05, | |
| "loss": 0.0433, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "eval_accuracy": 0.9829266169055779, | |
| "eval_f1": 0.8900519485573087, | |
| "eval_loss": 0.046879906207323074, | |
| "eval_precision": 0.902074459693958, | |
| "eval_recall": 0.8783456856495526, | |
| "eval_runtime": 75.3332, | |
| "eval_samples_per_second": 217.832, | |
| "eval_steps_per_second": 6.81, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.4937703141928494e-05, | |
| "loss": 0.04, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "eval_accuracy": 0.983201015273963, | |
| "eval_f1": 0.8977493905780533, | |
| "eval_loss": 0.04475250095129013, | |
| "eval_precision": 0.8837711139221668, | |
| "eval_recall": 0.9121769517299643, | |
| "eval_runtime": 73.7339, | |
| "eval_samples_per_second": 222.557, | |
| "eval_steps_per_second": 6.957, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 8.166305525460455e-06, | |
| "loss": 0.0365, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "eval_accuracy": 0.9843672083395997, | |
| "eval_f1": 0.9045125132863078, | |
| "eval_loss": 0.04261544719338417, | |
| "eval_precision": 0.9104419545596016, | |
| "eval_recall": 0.8986598056910257, | |
| "eval_runtime": 74.7052, | |
| "eval_samples_per_second": 219.663, | |
| "eval_steps_per_second": 6.867, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.394907908992416e-06, | |
| "loss": 0.0366, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "eval_accuracy": 0.9845123228613417, | |
| "eval_f1": 0.9050084368768215, | |
| "eval_loss": 0.041375741362571716, | |
| "eval_precision": 0.9037953352992991, | |
| "eval_recall": 0.9062247993548634, | |
| "eval_runtime": 75.1389, | |
| "eval_samples_per_second": 218.396, | |
| "eval_steps_per_second": 6.827, | |
| "step": 4000 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 4103, | |
| "num_train_epochs": 1, | |
| "save_steps": 500, | |
| "total_flos": 4183330761475008.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |