{ "best_metric": 0.8387653827667236, "best_model_checkpoint": "finetune_roberta-base/checkpoint-5476", "epoch": 2.0, "eval_steps": 500, "global_step": 5476, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.18, "learning_rate": 1.879474068663258e-05, "loss": 1.7148, "step": 500 }, { "epoch": 0.37, "learning_rate": 1.7577307036766496e-05, "loss": 1.075, "step": 1000 }, { "epoch": 0.55, "learning_rate": 1.6359873386900417e-05, "loss": 0.9886, "step": 1500 }, { "epoch": 0.73, "learning_rate": 1.5142439737034334e-05, "loss": 0.9376, "step": 2000 }, { "epoch": 0.91, "learning_rate": 1.3925006087168251e-05, "loss": 0.9256, "step": 2500 }, { "epoch": 1.0, "eval_loss": 0.8695176839828491, "eval_runtime": 78.4582, "eval_samples_per_second": 134.721, "eval_steps_per_second": 1.058, "step": 2738 }, { "epoch": 1.1, "learning_rate": 1.2707572437302169e-05, "loss": 0.8303, "step": 3000 }, { "epoch": 1.28, "learning_rate": 1.1490138787436084e-05, "loss": 0.7641, "step": 3500 }, { "epoch": 1.46, "learning_rate": 1.0275140004869735e-05, "loss": 0.7347, "step": 4000 }, { "epoch": 1.64, "learning_rate": 9.060141222303384e-06, "loss": 0.737, "step": 4500 }, { "epoch": 1.83, "learning_rate": 7.842707572437303e-06, "loss": 0.7466, "step": 5000 }, { "epoch": 2.0, "eval_loss": 0.8387653827667236, "eval_runtime": 77.9854, "eval_samples_per_second": 135.538, "eval_steps_per_second": 1.064, "step": 5476 } ], "logging_steps": 500, "max_steps": 8214, "num_train_epochs": 3, "save_steps": 500, "total_flos": 3.4334001889975296e+16, "trial_name": null, "trial_params": null }