| { | |
| "best_metric": 0.40056768847137303, | |
| "best_model_checkpoint": "distilbert-base-uncased-finetuned-cola/run-6/checkpoint-2140", | |
| "epoch": 4.0, | |
| "eval_steps": 500, | |
| "global_step": 2140, | |
| "is_hyper_param_search": true, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.5070736020231148e-06, | |
| "loss": 0.5992, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_loss": 0.5636755228042603, | |
| "eval_matthews_correlation": 0.0463559874942472, | |
| "eval_runtime": 0.7247, | |
| "eval_samples_per_second": 1439.132, | |
| "eval_steps_per_second": 91.067, | |
| "step": 535 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 1.047599942869726e-06, | |
| "loss": 0.5321, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_loss": 0.5432226061820984, | |
| "eval_matthews_correlation": 0.3132088164238629, | |
| "eval_runtime": 0.8222, | |
| "eval_samples_per_second": 1268.565, | |
| "eval_steps_per_second": 80.274, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 5.881262837163374e-07, | |
| "loss": 0.4898, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_loss": 0.527585506439209, | |
| "eval_matthews_correlation": 0.3945360499798347, | |
| "eval_runtime": 1.1962, | |
| "eval_samples_per_second": 871.948, | |
| "eval_steps_per_second": 55.176, | |
| "step": 1605 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "learning_rate": 1.286526245629488e-07, | |
| "loss": 0.4758, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_loss": 0.5270442366600037, | |
| "eval_matthews_correlation": 0.40056768847137303, | |
| "eval_runtime": 1.1465, | |
| "eval_samples_per_second": 909.688, | |
| "eval_steps_per_second": 57.564, | |
| "step": 2140 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 2140, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 4, | |
| "save_steps": 500, | |
| "total_flos": 173898906114156.0, | |
| "train_batch_size": 16, | |
| "trial_name": null, | |
| "trial_params": { | |
| "learning_rate": 1.9665472611765034e-06, | |
| "num_train_epochs": 4, | |
| "per_device_train_batch_size": 16, | |
| "seed": 17 | |
| } | |
| } | |