| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.19184652278177458, | |
| "global_step": 200, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.995201535508638e-05, | |
| "loss": 0.6593, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.990403071017275e-05, | |
| "loss": 0.5495, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.985604606525912e-05, | |
| "loss": 0.507, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.980806142034548e-05, | |
| "loss": 0.4512, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.976007677543187e-05, | |
| "loss": 0.417, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.971209213051824e-05, | |
| "loss": 0.4129, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.96641074856046e-05, | |
| "loss": 0.4053, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.961612284069098e-05, | |
| "loss": 0.3658, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.956813819577736e-05, | |
| "loss": 0.3779, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.952015355086372e-05, | |
| "loss": 0.3791, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.94721689059501e-05, | |
| "loss": 0.3492, | |
| "step": 55 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.942418426103647e-05, | |
| "loss": 0.341, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.937619961612284e-05, | |
| "loss": 0.3388, | |
| "step": 65 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.932821497120922e-05, | |
| "loss": 0.3597, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.928023032629559e-05, | |
| "loss": 0.3356, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.923224568138196e-05, | |
| "loss": 0.3317, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.918426103646834e-05, | |
| "loss": 0.3424, | |
| "step": 85 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.913627639155471e-05, | |
| "loss": 0.333, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.908829174664108e-05, | |
| "loss": 0.3145, | |
| "step": 95 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.904030710172744e-05, | |
| "loss": 0.3213, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.899232245681382e-05, | |
| "loss": 0.3139, | |
| "step": 105 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.89443378119002e-05, | |
| "loss": 0.3155, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.889635316698656e-05, | |
| "loss": 0.3219, | |
| "step": 115 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.884836852207294e-05, | |
| "loss": 0.3049, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.880038387715931e-05, | |
| "loss": 0.3171, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.875239923224568e-05, | |
| "loss": 0.2836, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.870441458733206e-05, | |
| "loss": 0.3102, | |
| "step": 135 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.865642994241843e-05, | |
| "loss": 0.2853, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.86084452975048e-05, | |
| "loss": 0.3273, | |
| "step": 145 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.856046065259118e-05, | |
| "loss": 0.3093, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.851247600767755e-05, | |
| "loss": 0.3112, | |
| "step": 155 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.846449136276392e-05, | |
| "loss": 0.308, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.841650671785028e-05, | |
| "loss": 0.3217, | |
| "step": 165 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.836852207293667e-05, | |
| "loss": 0.2773, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.832053742802304e-05, | |
| "loss": 0.2932, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.82725527831094e-05, | |
| "loss": 0.3055, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.822456813819578e-05, | |
| "loss": 0.282, | |
| "step": 185 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.817658349328216e-05, | |
| "loss": 0.2682, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.812859884836852e-05, | |
| "loss": 0.2913, | |
| "step": 195 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.80806142034549e-05, | |
| "loss": 0.2813, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "eval_accuracy": 0.6656578015326164, | |
| "eval_accuracy_sklearn": 0.6656578015326164, | |
| "eval_f1": 0.6489045171760888, | |
| "eval_loss": 0.7392101883888245, | |
| "eval_precision": 0.809106239460371, | |
| "eval_recall": 0.5416572589749379, | |
| "eval_runtime": 84.7791, | |
| "eval_samples_per_second": 183.17, | |
| "eval_steps_per_second": 11.453, | |
| "step": 200 | |
| } | |
| ], | |
| "max_steps": 10420, | |
| "num_train_epochs": 10, | |
| "total_flos": 2.6942572068864e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |