| { | |
| "best_metric": 0.9402497791791053, | |
| "best_model_checkpoint": "model_saves/roberta-large_lemon_10k_1_p3/checkpoint-268", | |
| "epoch": 7.0, | |
| "global_step": 1876, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.9402497791791053, | |
| "eval_loss": 0.43265554308891296, | |
| "eval_runtime": 3.4386, | |
| "eval_samples_per_second": 1274.348, | |
| "eval_steps_per_second": 10.178, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.9400597054976018, | |
| "eval_loss": 0.4408949613571167, | |
| "eval_runtime": 3.436, | |
| "eval_samples_per_second": 1275.314, | |
| "eval_steps_per_second": 10.186, | |
| "step": 536 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9397019197441832, | |
| "eval_loss": 0.47041797637939453, | |
| "eval_runtime": 3.4489, | |
| "eval_samples_per_second": 1270.55, | |
| "eval_steps_per_second": 10.148, | |
| "step": 804 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 1e-05, | |
| "loss": 0.317, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.9388913113965943, | |
| "eval_loss": 0.5034393668174744, | |
| "eval_runtime": 3.4571, | |
| "eval_samples_per_second": 1267.527, | |
| "eval_steps_per_second": 10.124, | |
| "step": 1072 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.9389248538109773, | |
| "eval_loss": 0.5431345701217651, | |
| "eval_runtime": 3.4436, | |
| "eval_samples_per_second": 1272.493, | |
| "eval_steps_per_second": 10.164, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.9384161271928353, | |
| "eval_loss": 0.5829800367355347, | |
| "eval_runtime": 3.4558, | |
| "eval_samples_per_second": 1268.013, | |
| "eval_steps_per_second": 10.128, | |
| "step": 1608 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.938678876105502, | |
| "eval_loss": 0.650184690952301, | |
| "eval_runtime": 3.5028, | |
| "eval_samples_per_second": 1251.012, | |
| "eval_steps_per_second": 9.992, | |
| "step": 1876 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "step": 1876, | |
| "total_flos": 2.467935271727923e+16, | |
| "train_loss": 0.06943080877698561, | |
| "train_runtime": 261.0381, | |
| "train_samples_per_second": 1971.207, | |
| "train_steps_per_second": 15.4 | |
| } | |
| ], | |
| "max_steps": 4020, | |
| "num_train_epochs": 15, | |
| "total_flos": 2.467935271727923e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |