| { | |
| "best_metric": 0.6180649399757385, | |
| "best_model_checkpoint": "add_BERT_no_pretrain_cola/checkpoint-335", | |
| "epoch": 10.0, | |
| "global_step": 670, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 3.9200000000000004e-05, | |
| "loss": 0.6339, | |
| "step": 67 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.6912751793861389, | |
| "eval_loss": 0.6182041168212891, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 1.9602, | |
| "eval_samples_per_second": 532.087, | |
| "eval_steps_per_second": 4.591, | |
| "step": 67 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 3.8400000000000005e-05, | |
| "loss": 0.6177, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.6912751793861389, | |
| "eval_loss": 0.6420519351959229, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 1.9688, | |
| "eval_samples_per_second": 529.756, | |
| "eval_steps_per_second": 4.571, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 3.76e-05, | |
| "loss": 0.6204, | |
| "step": 201 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.6912751793861389, | |
| "eval_loss": 0.6295004487037659, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 1.961, | |
| "eval_samples_per_second": 531.871, | |
| "eval_steps_per_second": 4.589, | |
| "step": 201 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 3.680000000000001e-05, | |
| "loss": 0.6182, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.6912751793861389, | |
| "eval_loss": 0.6267786026000977, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 1.9603, | |
| "eval_samples_per_second": 532.067, | |
| "eval_steps_per_second": 4.591, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 3.6e-05, | |
| "loss": 0.6149, | |
| "step": 335 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.6912751793861389, | |
| "eval_loss": 0.6180649399757385, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 1.9731, | |
| "eval_samples_per_second": 528.608, | |
| "eval_steps_per_second": 4.561, | |
| "step": 335 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 3.52e-05, | |
| "loss": 0.612, | |
| "step": 402 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.6912751793861389, | |
| "eval_loss": 0.6188631057739258, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 1.9662, | |
| "eval_samples_per_second": 530.452, | |
| "eval_steps_per_second": 4.577, | |
| "step": 402 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 3.44e-05, | |
| "loss": 0.6132, | |
| "step": 469 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.6912751793861389, | |
| "eval_loss": 0.6292060017585754, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 1.9618, | |
| "eval_samples_per_second": 531.664, | |
| "eval_steps_per_second": 4.588, | |
| "step": 469 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 3.3600000000000004e-05, | |
| "loss": 0.6125, | |
| "step": 536 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.6912751793861389, | |
| "eval_loss": 0.6185278296470642, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 1.9633, | |
| "eval_samples_per_second": 531.253, | |
| "eval_steps_per_second": 4.584, | |
| "step": 536 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "learning_rate": 3.28e-05, | |
| "loss": 0.6108, | |
| "step": 603 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.6912751793861389, | |
| "eval_loss": 0.6280407309532166, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 1.9674, | |
| "eval_samples_per_second": 530.136, | |
| "eval_steps_per_second": 4.575, | |
| "step": 603 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "learning_rate": 3.2000000000000005e-05, | |
| "loss": 0.6092, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_accuracy": 0.6912751793861389, | |
| "eval_loss": 0.6309932470321655, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 1.9598, | |
| "eval_samples_per_second": 532.187, | |
| "eval_steps_per_second": 4.592, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "step": 670, | |
| "total_flos": 1.221287775043584e+16, | |
| "train_loss": 0.6162634493699715, | |
| "train_runtime": 600.8403, | |
| "train_samples_per_second": 711.587, | |
| "train_steps_per_second": 5.576 | |
| } | |
| ], | |
| "max_steps": 3350, | |
| "num_train_epochs": 50, | |
| "total_flos": 1.221287775043584e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |