| { | |
| "best_global_step": 30, | |
| "best_metric": 0.6030584573745728, | |
| "best_model_checkpoint": "bert_base_rand_50_v1_mrpc/checkpoint-30", | |
| "epoch": 7.0, | |
| "eval_steps": 500, | |
| "global_step": 105, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "grad_norm": 1.0738521814346313, | |
| "learning_rate": 4.9e-05, | |
| "loss": 0.6274, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.6740196078431373, | |
| "eval_combined_score": 0.7278144016227182, | |
| "eval_f1": 0.7816091954022989, | |
| "eval_loss": 0.6055310368537903, | |
| "eval_runtime": 0.2767, | |
| "eval_samples_per_second": 1474.47, | |
| "eval_steps_per_second": 7.228, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 1.3434077501296997, | |
| "learning_rate": 4.8e-05, | |
| "loss": 0.57, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.7009803921568627, | |
| "eval_combined_score": 0.7438468394350747, | |
| "eval_f1": 0.7867132867132867, | |
| "eval_loss": 0.6030584573745728, | |
| "eval_runtime": 0.2694, | |
| "eval_samples_per_second": 1514.212, | |
| "eval_steps_per_second": 7.423, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 2.2089107036590576, | |
| "learning_rate": 4.7e-05, | |
| "loss": 0.4881, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.6862745098039216, | |
| "eval_combined_score": 0.7267736185383245, | |
| "eval_f1": 0.7672727272727272, | |
| "eval_loss": 0.6268609762191772, | |
| "eval_runtime": 0.2758, | |
| "eval_samples_per_second": 1479.07, | |
| "eval_steps_per_second": 7.25, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 4.52411413192749, | |
| "learning_rate": 4.600000000000001e-05, | |
| "loss": 0.3638, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.6544117647058824, | |
| "eval_combined_score": 0.6968916494509079, | |
| "eval_f1": 0.7393715341959335, | |
| "eval_loss": 0.7924944758415222, | |
| "eval_runtime": 0.2711, | |
| "eval_samples_per_second": 1504.858, | |
| "eval_steps_per_second": 7.377, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 3.8104875087738037, | |
| "learning_rate": 4.5e-05, | |
| "loss": 0.2309, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.6666666666666666, | |
| "eval_combined_score": 0.7136150234741784, | |
| "eval_f1": 0.7605633802816901, | |
| "eval_loss": 0.9224503636360168, | |
| "eval_runtime": 0.2729, | |
| "eval_samples_per_second": 1495.255, | |
| "eval_steps_per_second": 7.33, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 4.772459030151367, | |
| "learning_rate": 4.4000000000000006e-05, | |
| "loss": 0.1597, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.678921568627451, | |
| "eval_combined_score": 0.7247497510387693, | |
| "eval_f1": 0.7705779334500875, | |
| "eval_loss": 1.031596064567566, | |
| "eval_runtime": 0.2738, | |
| "eval_samples_per_second": 1490.377, | |
| "eval_steps_per_second": 7.306, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 5.311474800109863, | |
| "learning_rate": 4.3e-05, | |
| "loss": 0.0979, | |
| "step": 105 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.6470588235294118, | |
| "eval_combined_score": 0.6985294117647058, | |
| "eval_f1": 0.75, | |
| "eval_loss": 1.3662556409835815, | |
| "eval_runtime": 0.2832, | |
| "eval_samples_per_second": 1440.776, | |
| "eval_steps_per_second": 7.063, | |
| "step": 105 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "step": 105, | |
| "total_flos": 3377819728711680.0, | |
| "train_loss": 0.3625380856650216, | |
| "train_runtime": 57.9035, | |
| "train_samples_per_second": 3167.341, | |
| "train_steps_per_second": 12.953 | |
| } | |
| ], | |
| "logging_steps": 1, | |
| "max_steps": 750, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 50, | |
| "save_steps": 500, | |
| "stateful_callbacks": { | |
| "EarlyStoppingCallback": { | |
| "args": { | |
| "early_stopping_patience": 5, | |
| "early_stopping_threshold": 0.0 | |
| }, | |
| "attributes": { | |
| "early_stopping_patience_counter": 5 | |
| } | |
| }, | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": true | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 3377819728711680.0, | |
| "train_batch_size": 256, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |