| { | |
| "best_metric": 0.940171513545545, | |
| "best_model_checkpoint": "model_saves/roberta-large_lemon_10k_2_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": 0.940171513545545, | |
| "eval_loss": 0.43810606002807617, | |
| "eval_runtime": 3.415, | |
| "eval_samples_per_second": 1283.177, | |
| "eval_steps_per_second": 10.249, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.9395845212938427, | |
| "eval_loss": 0.449770987033844, | |
| "eval_runtime": 3.4195, | |
| "eval_samples_per_second": 1281.488, | |
| "eval_steps_per_second": 10.236, | |
| "step": 536 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9389975290421404, | |
| "eval_loss": 0.476426899433136, | |
| "eval_runtime": 3.4072, | |
| "eval_samples_per_second": 1286.11, | |
| "eval_steps_per_second": 10.272, | |
| "step": 804 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 1e-05, | |
| "loss": 0.2859, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.9391372891020696, | |
| "eval_loss": 0.5197746157646179, | |
| "eval_runtime": 3.4116, | |
| "eval_samples_per_second": 1284.432, | |
| "eval_steps_per_second": 10.259, | |
| "step": 1072 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.9385950200695446, | |
| "eval_loss": 0.5668849349021912, | |
| "eval_runtime": 3.406, | |
| "eval_samples_per_second": 1286.537, | |
| "eval_steps_per_second": 10.276, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.9381981014993459, | |
| "eval_loss": 0.6483766436576843, | |
| "eval_runtime": 3.4058, | |
| "eval_samples_per_second": 1286.611, | |
| "eval_steps_per_second": 10.276, | |
| "step": 1608 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.9379521237938707, | |
| "eval_loss": 0.6937921643257141, | |
| "eval_runtime": 3.4208, | |
| "eval_samples_per_second": 1280.993, | |
| "eval_steps_per_second": 10.232, | |
| "step": 1876 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "step": 1876, | |
| "total_flos": 2.4706569119727616e+16, | |
| "train_loss": 0.06039006166112448, | |
| "train_runtime": 257.6645, | |
| "train_samples_per_second": 1997.015, | |
| "train_steps_per_second": 15.602 | |
| } | |
| ], | |
| "max_steps": 4020, | |
| "num_train_epochs": 15, | |
| "total_flos": 2.4706569119727616e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |