| { | |
| "best_metric": 0.13342921435832977, | |
| "best_model_checkpoint": "/home/dongyeop/owan0002/revision-preference/code/preference_classifier/coedit_t5/output_r1_iter_wo_p/checkpoint-405", | |
| "epoch": 20.0, | |
| "global_step": 540, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "eval_bleu": 0.0, | |
| "eval_gen_len": 2.9953, | |
| "eval_loss": 0.2727777659893036, | |
| "eval_runtime": 10.2846, | |
| "eval_samples_per_second": 164.517, | |
| "eval_steps_per_second": 1.361, | |
| "step": 27 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_bleu": 0.0, | |
| "eval_gen_len": 2.6791, | |
| "eval_loss": 0.26499590277671814, | |
| "eval_runtime": 9.934, | |
| "eval_samples_per_second": 170.324, | |
| "eval_steps_per_second": 1.409, | |
| "step": 54 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_bleu": 0.0, | |
| "eval_gen_len": 2.1874, | |
| "eval_loss": 0.2637251317501068, | |
| "eval_runtime": 9.8272, | |
| "eval_samples_per_second": 172.175, | |
| "eval_steps_per_second": 1.425, | |
| "step": 81 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_bleu": 0.0, | |
| "eval_gen_len": 2.2973, | |
| "eval_loss": 0.24176451563835144, | |
| "eval_runtime": 9.7849, | |
| "eval_samples_per_second": 172.92, | |
| "eval_steps_per_second": 1.431, | |
| "step": 108 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_bleu": 0.0, | |
| "eval_gen_len": 2.2494, | |
| "eval_loss": 0.2737601399421692, | |
| "eval_runtime": 10.431, | |
| "eval_samples_per_second": 162.209, | |
| "eval_steps_per_second": 1.342, | |
| "step": 135 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_bleu": 0.0, | |
| "eval_gen_len": 2.3812, | |
| "eval_loss": 0.19136066734790802, | |
| "eval_runtime": 9.9867, | |
| "eval_samples_per_second": 169.425, | |
| "eval_steps_per_second": 1.402, | |
| "step": 162 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_bleu": 0.0, | |
| "eval_gen_len": 2.3983, | |
| "eval_loss": 0.16406123340129852, | |
| "eval_runtime": 9.9234, | |
| "eval_samples_per_second": 170.505, | |
| "eval_steps_per_second": 1.411, | |
| "step": 189 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_bleu": 0.0, | |
| "eval_gen_len": 2.3995, | |
| "eval_loss": 0.16948193311691284, | |
| "eval_runtime": 10.0828, | |
| "eval_samples_per_second": 167.811, | |
| "eval_steps_per_second": 1.389, | |
| "step": 216 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_bleu": 0.0, | |
| "eval_gen_len": 2.4167, | |
| "eval_loss": 0.15213139355182648, | |
| "eval_runtime": 9.7652, | |
| "eval_samples_per_second": 173.268, | |
| "eval_steps_per_second": 1.434, | |
| "step": 243 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_bleu": 0.0, | |
| "eval_gen_len": 2.4167, | |
| "eval_loss": 0.15690775215625763, | |
| "eval_runtime": 9.8519, | |
| "eval_samples_per_second": 171.743, | |
| "eval_steps_per_second": 1.421, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_bleu": 0.0, | |
| "eval_gen_len": 2.4137, | |
| "eval_loss": 0.16145527362823486, | |
| "eval_runtime": 9.771, | |
| "eval_samples_per_second": 173.166, | |
| "eval_steps_per_second": 1.433, | |
| "step": 297 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_bleu": 0.0, | |
| "eval_gen_len": 2.4238, | |
| "eval_loss": 0.1473228484392166, | |
| "eval_runtime": 9.9975, | |
| "eval_samples_per_second": 169.242, | |
| "eval_steps_per_second": 1.4, | |
| "step": 324 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "eval_bleu": 0.0, | |
| "eval_gen_len": 2.4255, | |
| "eval_loss": 0.13761219382286072, | |
| "eval_runtime": 9.8987, | |
| "eval_samples_per_second": 170.931, | |
| "eval_steps_per_second": 1.414, | |
| "step": 351 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "eval_bleu": 0.0, | |
| "eval_gen_len": 2.419, | |
| "eval_loss": 0.14949482679367065, | |
| "eval_runtime": 10.1968, | |
| "eval_samples_per_second": 165.935, | |
| "eval_steps_per_second": 1.373, | |
| "step": 378 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "eval_bleu": 0.0, | |
| "eval_gen_len": 2.432, | |
| "eval_loss": 0.13342921435832977, | |
| "eval_runtime": 9.9102, | |
| "eval_samples_per_second": 170.734, | |
| "eval_steps_per_second": 1.413, | |
| "step": 405 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "eval_bleu": 0.0, | |
| "eval_gen_len": 2.4214, | |
| "eval_loss": 0.1473984271287918, | |
| "eval_runtime": 9.8919, | |
| "eval_samples_per_second": 171.049, | |
| "eval_steps_per_second": 1.415, | |
| "step": 432 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "eval_bleu": 0.0, | |
| "eval_gen_len": 2.4291, | |
| "eval_loss": 0.1483658105134964, | |
| "eval_runtime": 9.911, | |
| "eval_samples_per_second": 170.72, | |
| "eval_steps_per_second": 1.413, | |
| "step": 459 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "eval_bleu": 0.0, | |
| "eval_gen_len": 2.4297, | |
| "eval_loss": 0.1407216191291809, | |
| "eval_runtime": 9.8716, | |
| "eval_samples_per_second": 171.4, | |
| "eval_steps_per_second": 1.418, | |
| "step": 486 | |
| }, | |
| { | |
| "epoch": 18.52, | |
| "learning_rate": 0.0001, | |
| "loss": 0.1905, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 19.0, | |
| "eval_bleu": 0.0, | |
| "eval_gen_len": 2.4208, | |
| "eval_loss": 0.15682467818260193, | |
| "eval_runtime": 9.8237, | |
| "eval_samples_per_second": 172.237, | |
| "eval_steps_per_second": 1.425, | |
| "step": 513 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "eval_bleu": 0.0, | |
| "eval_gen_len": 2.4261, | |
| "eval_loss": 0.16313956677913666, | |
| "eval_runtime": 9.8505, | |
| "eval_samples_per_second": 171.767, | |
| "eval_steps_per_second": 1.421, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "step": 540, | |
| "total_flos": 2.4870877697304166e+17, | |
| "train_loss": 0.18235577079984877, | |
| "train_runtime": 3852.2155, | |
| "train_samples_per_second": 71.143, | |
| "train_steps_per_second": 0.14 | |
| } | |
| ], | |
| "max_steps": 540, | |
| "num_train_epochs": 20, | |
| "total_flos": 2.4870877697304166e+17, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |