| { | |
| "best_metric": 0.6931639313697815, | |
| "best_model_checkpoint": "checkpoints/electra_26_4_1/checkpoint-13858", | |
| "epoch": 13.0, | |
| "global_step": 13858, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 3.904917901570934e-05, | |
| "loss": 0.7426, | |
| "step": 1066 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_auc_micro": 0.5212763137512771, | |
| "eval_loss": 0.6989310383796692, | |
| "eval_runtime": 23.3554, | |
| "eval_samples_per_second": 128.835, | |
| "eval_steps_per_second": 8.092, | |
| "step": 1066 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 3.5795080764400225e-05, | |
| "loss": 0.6964, | |
| "step": 2132 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_auc_micro": 0.5228480764195904, | |
| "eval_loss": 0.6949135661125183, | |
| "eval_runtime": 23.3685, | |
| "eval_samples_per_second": 128.763, | |
| "eval_steps_per_second": 8.088, | |
| "step": 2132 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 3.254098251309111e-05, | |
| "loss": 0.6944, | |
| "step": 3198 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_auc_micro": 0.5232832697413072, | |
| "eval_loss": 0.6939617395401001, | |
| "eval_runtime": 23.1376, | |
| "eval_samples_per_second": 130.048, | |
| "eval_steps_per_second": 8.169, | |
| "step": 3198 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 2.9286884261782e-05, | |
| "loss": 0.6938, | |
| "step": 4264 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_auc_micro": 0.5234944214364559, | |
| "eval_loss": 0.6935874223709106, | |
| "eval_runtime": 23.1396, | |
| "eval_samples_per_second": 130.037, | |
| "eval_steps_per_second": 8.168, | |
| "step": 4264 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 2.603278601047289e-05, | |
| "loss": 0.6935, | |
| "step": 5330 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_auc_micro": 0.5237105566327973, | |
| "eval_loss": 0.6934056282043457, | |
| "eval_runtime": 23.0585, | |
| "eval_samples_per_second": 130.494, | |
| "eval_steps_per_second": 8.197, | |
| "step": 5330 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 2.2778687759163776e-05, | |
| "loss": 0.6934, | |
| "step": 6396 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_auc_micro": 0.5238991498421133, | |
| "eval_loss": 0.6933064460754395, | |
| "eval_runtime": 23.1864, | |
| "eval_samples_per_second": 129.775, | |
| "eval_steps_per_second": 8.151, | |
| "step": 6396 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 1.952458950785467e-05, | |
| "loss": 0.6933, | |
| "step": 7462 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_auc_micro": 0.5239438807088983, | |
| "eval_loss": 0.6932488679885864, | |
| "eval_runtime": 23.1343, | |
| "eval_samples_per_second": 130.067, | |
| "eval_steps_per_second": 8.17, | |
| "step": 7462 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 1.6270491256545556e-05, | |
| "loss": 0.6932, | |
| "step": 8528 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_auc_micro": 0.523929452444219, | |
| "eval_loss": 0.6932142972946167, | |
| "eval_runtime": 23.2347, | |
| "eval_samples_per_second": 129.505, | |
| "eval_steps_per_second": 8.134, | |
| "step": 8528 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "learning_rate": 1.3016393005236445e-05, | |
| "loss": 0.6932, | |
| "step": 9594 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_auc_micro": 0.5239810749738886, | |
| "eval_loss": 0.6931925415992737, | |
| "eval_runtime": 23.1045, | |
| "eval_samples_per_second": 130.234, | |
| "eval_steps_per_second": 8.18, | |
| "step": 9594 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "learning_rate": 9.762294753927335e-06, | |
| "loss": 0.6932, | |
| "step": 10660 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_auc_micro": 0.5241205756951521, | |
| "eval_loss": 0.6931790709495544, | |
| "eval_runtime": 23.1973, | |
| "eval_samples_per_second": 129.713, | |
| "eval_steps_per_second": 8.148, | |
| "step": 10660 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "learning_rate": 6.508196502618222e-06, | |
| "loss": 0.6932, | |
| "step": 11726 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_auc_micro": 0.5241602931373088, | |
| "eval_loss": 0.6931705474853516, | |
| "eval_runtime": 23.0891, | |
| "eval_samples_per_second": 130.321, | |
| "eval_steps_per_second": 8.186, | |
| "step": 11726 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "learning_rate": 3.254098251309111e-06, | |
| "loss": 0.6932, | |
| "step": 12792 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_auc_micro": 0.5242074068777832, | |
| "eval_loss": 0.6931658387184143, | |
| "eval_runtime": 23.1506, | |
| "eval_samples_per_second": 129.975, | |
| "eval_steps_per_second": 8.164, | |
| "step": 12792 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "learning_rate": 0.0, | |
| "loss": 0.6932, | |
| "step": 13858 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "eval_auc_micro": 0.5243155309888969, | |
| "eval_loss": 0.6931639313697815, | |
| "eval_runtime": 23.1954, | |
| "eval_samples_per_second": 129.724, | |
| "eval_steps_per_second": 8.148, | |
| "step": 13858 | |
| } | |
| ], | |
| "max_steps": 13858, | |
| "num_train_epochs": 13, | |
| "total_flos": 0.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |