{ "best_metric": 0.45749443397572137, "best_model_checkpoint": "distilbert-base-uncased-finetuned-cola/run-2/checkpoint-2675", "epoch": 5.0, "eval_steps": 500, "global_step": 2675, "is_hyper_param_search": true, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.9345794392523364, "grad_norm": 5.135624885559082, "learning_rate": 3.571015099578109e-06, "loss": 0.5612, "step": 500 }, { "epoch": 1.0, "eval_loss": 0.5108451843261719, "eval_matthews_correlation": 0.4068953869702192, "eval_runtime": 0.3429, "eval_samples_per_second": 3041.66, "eval_steps_per_second": 192.473, "step": 535 }, { "epoch": 1.8691588785046729, "grad_norm": 11.39373779296875, "learning_rate": 2.750092088180842e-06, "loss": 0.4687, "step": 1000 }, { "epoch": 2.0, "eval_loss": 0.5220178365707397, "eval_matthews_correlation": 0.415128776020243, "eval_runtime": 0.3609, "eval_samples_per_second": 2889.801, "eval_steps_per_second": 182.864, "step": 1070 }, { "epoch": 2.803738317757009, "grad_norm": 14.82923412322998, "learning_rate": 1.929169076783576e-06, "loss": 0.4153, "step": 1500 }, { "epoch": 3.0, "eval_loss": 0.5056647658348083, "eval_matthews_correlation": 0.45042204436320565, "eval_runtime": 0.396, "eval_samples_per_second": 2633.94, "eval_steps_per_second": 166.673, "step": 1605 }, { "epoch": 3.7383177570093458, "grad_norm": 9.636689186096191, "learning_rate": 1.1082460653863094e-06, "loss": 0.3805, "step": 2000 }, { "epoch": 4.0, "eval_loss": 0.5095609426498413, "eval_matthews_correlation": 0.4504043574931814, "eval_runtime": 0.3683, "eval_samples_per_second": 2832.059, "eval_steps_per_second": 179.21, "step": 2140 }, { "epoch": 4.672897196261682, "grad_norm": 6.002148628234863, "learning_rate": 2.8732305398904323e-07, "loss": 0.358, "step": 2500 }, { "epoch": 5.0, "eval_loss": 0.525900661945343, "eval_matthews_correlation": 0.45749443397572137, "eval_runtime": 0.3551, "eval_samples_per_second": 2936.996, "eval_steps_per_second": 185.85, "step": 2675 } ], "logging_steps": 500, "max_steps": 2675, "num_input_tokens_seen": 0, "num_train_epochs": 5, "save_steps": 500, "total_flos": 214443502942248.0, "train_batch_size": 16, "trial_name": null, "trial_params": { "learning_rate": 4.391938110975375e-06, "num_train_epochs": 5, "per_device_train_batch_size": 16, "seed": 23 } }