{ "best_metric": 0.9268057650410592, "best_model_checkpoint": "FS_OUT/harish/PT-UP-xlmR-FalseTrue-0_0_BEST/FalseTrue-0/0/checkpoint-35", "epoch": 7.0, "global_step": 35, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 1.0, "eval_accuracy": 0.7582417726516724, "eval_f1": 0.7521978021978022, "eval_loss": 0.6778414845466614, "eval_runtime": 0.7947, "eval_samples_per_second": 343.537, "eval_steps_per_second": 44.043, "step": 5 }, { "epoch": 2.0, "eval_accuracy": 0.8571428656578064, "eval_f1": 0.8575954937511002, "eval_loss": 0.3829486668109894, "eval_runtime": 0.7941, "eval_samples_per_second": 343.785, "eval_steps_per_second": 44.075, "step": 10 }, { "epoch": 3.0, "eval_accuracy": 0.9194139242172241, "eval_f1": 0.9195497484971169, "eval_loss": 0.32353681325912476, "eval_runtime": 0.8016, "eval_samples_per_second": 340.571, "eval_steps_per_second": 43.663, "step": 15 }, { "epoch": 4.0, "eval_accuracy": 0.9157509207725525, "eval_f1": 0.9156195017041976, "eval_loss": 0.2483082264661789, "eval_runtime": 0.7936, "eval_samples_per_second": 343.995, "eval_steps_per_second": 44.102, "step": 20 }, { "epoch": 5.0, "eval_accuracy": 0.8937729001045227, "eval_f1": 0.8942076821925489, "eval_loss": 0.30191415548324585, "eval_runtime": 0.7926, "eval_samples_per_second": 344.419, "eval_steps_per_second": 44.156, "step": 25 }, { "epoch": 6.0, "eval_accuracy": 0.9120879173278809, "eval_f1": 0.9123452157598498, "eval_loss": 0.24024997651576996, "eval_runtime": 0.7937, "eval_samples_per_second": 343.965, "eval_steps_per_second": 44.098, "step": 30 }, { "epoch": 7.0, "eval_accuracy": 0.9267399311065674, "eval_f1": 0.9268057650410592, "eval_loss": 0.21064965426921844, "eval_runtime": 0.795, "eval_samples_per_second": 343.409, "eval_steps_per_second": 44.027, "step": 35 } ], "max_steps": 45, "num_train_epochs": 9, "total_flos": 233184263510016.0, "trial_name": null, "trial_params": null }