| { |
| "best_metric": 0.6532846715328468, |
| "best_model_checkpoint": "./outputs/en_es_multilingual/mdeberta-v3-base-output/met_xnli_test_en_8_0.01_0.00005_11-17-23_09-47/checkpoint-3630", |
| "epoch": 4.0, |
| "global_step": 7260, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 1.0, |
| "learning_rate": 3.989595545134818e-05, |
| "loss": 0.0887, |
| "step": 1815 |
| }, |
| { |
| "epoch": 1.0, |
| "eval_accuracy": 0.9858611825192802, |
| "eval_f1": 0.6258389261744967, |
| "eval_loss": 0.04259440675377846, |
| "eval_precision": 0.6109746109746109, |
| "eval_recall": 0.641444539982803, |
| "eval_runtime": 14.5185, |
| "eval_samples_per_second": 334.951, |
| "eval_steps_per_second": 41.877, |
| "step": 1815 |
| }, |
| { |
| "epoch": 2.0, |
| "learning_rate": 2.6597303634232123e-05, |
| "loss": 0.0228, |
| "step": 3630 |
| }, |
| { |
| "epoch": 2.0, |
| "eval_accuracy": 0.9879240851820115, |
| "eval_f1": 0.6532846715328468, |
| "eval_loss": 0.034684374928474426, |
| "eval_precision": 0.6958211856171039, |
| "eval_recall": 0.6156491831470335, |
| "eval_runtime": 13.7183, |
| "eval_samples_per_second": 354.489, |
| "eval_steps_per_second": 44.32, |
| "step": 3630 |
| }, |
| { |
| "epoch": 3.0, |
| "learning_rate": 1.3298651817116062e-05, |
| "loss": 0.0104, |
| "step": 5445 |
| }, |
| { |
| "epoch": 3.0, |
| "eval_accuracy": 0.9885270874988099, |
| "eval_f1": 0.6226415094339622, |
| "eval_loss": 0.05654400959610939, |
| "eval_precision": 0.7973154362416107, |
| "eval_recall": 0.5107480653482374, |
| "eval_runtime": 12.5259, |
| "eval_samples_per_second": 388.237, |
| "eval_steps_per_second": 48.54, |
| "step": 5445 |
| }, |
| { |
| "epoch": 4.0, |
| "learning_rate": 0.0, |
| "loss": 0.0046, |
| "step": 7260 |
| }, |
| { |
| "epoch": 4.0, |
| "eval_accuracy": 0.9881779808943476, |
| "eval_f1": 0.6463007159904535, |
| "eval_loss": 0.05788963660597801, |
| "eval_precision": 0.7263948497854077, |
| "eval_recall": 0.5821152192605331, |
| "eval_runtime": 13.3972, |
| "eval_samples_per_second": 362.986, |
| "eval_steps_per_second": 45.383, |
| "step": 7260 |
| }, |
| { |
| "epoch": 4.0, |
| "step": 7260, |
| "total_flos": 1062049033953072.0, |
| "train_loss": 0.031637020347532165, |
| "train_runtime": 972.4399, |
| "train_samples_per_second": 59.722, |
| "train_steps_per_second": 7.466 |
| } |
| ], |
| "max_steps": 7260, |
| "num_train_epochs": 4, |
| "total_flos": 1062049033953072.0, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
