{ "best_metric": null, "best_model_checkpoint": null, "epoch": 6.991735537190083, "global_step": 420, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 1.66, "learning_rate": 7.619047619047618e-05, "loss": 1.2894, "step": 100 }, { "epoch": 1.66, "eval_accuracy": 0.4958677589893341, "eval_loss": 1.2452681064605713, "eval_runtime": 7.5077, "eval_samples_per_second": 16.117, "eval_steps_per_second": 4.129, "step": 100 }, { "epoch": 3.33, "learning_rate": 5.285714285714286e-05, "loss": 0.7978, "step": 200 }, { "epoch": 3.33, "eval_accuracy": 0.7768595218658447, "eval_loss": 0.746031641960144, "eval_runtime": 7.1743, "eval_samples_per_second": 16.866, "eval_steps_per_second": 4.321, "step": 200 }, { "epoch": 4.99, "learning_rate": 2.9047619047619052e-05, "loss": 0.4372, "step": 300 }, { "epoch": 4.99, "eval_accuracy": 0.9008264541625977, "eval_loss": 0.3735601603984833, "eval_runtime": 7.2008, "eval_samples_per_second": 16.804, "eval_steps_per_second": 4.305, "step": 300 }, { "epoch": 6.66, "learning_rate": 5.4761904761904765e-06, "loss": 0.246, "step": 400 }, { "epoch": 6.66, "eval_accuracy": 0.9504132270812988, "eval_loss": 0.24308893084526062, "eval_runtime": 10.0989, "eval_samples_per_second": 11.982, "eval_steps_per_second": 3.07, "step": 400 }, { "epoch": 6.99, "step": 420, "total_flos": 1.8354201823649805e+17, "train_loss": 0.6664659727187384, "train_runtime": 440.1582, "train_samples_per_second": 7.681, "train_steps_per_second": 0.954 } ], "max_steps": 420, "num_train_epochs": 7, "total_flos": 1.8354201823649805e+17, "trial_name": null, "trial_params": null }