ModelTC/bart-base-cola

1{2  "best_metric": null,3  "best_model_checkpoint": null,4  "epoch": 8.208955223880597,5  "global_step": 2200,6  "is_hyper_param_search": false,7  "is_local_process_zero": true,8  "is_world_process_zero": true,9  "log_history": [10    {11      "epoch": 0.75,12      "learning_rate": 3.92749244712991e-05,13      "loss": 0.5968,14      "step": 20015    },16    {17      "epoch": 0.75,18      "eval_loss": 0.5482025742530823,19      "eval_matthews_correlation": 0.4044551659076911,20      "eval_runtime": 7.0014,21      "eval_samples_per_second": 148.97,22      "eval_steps_per_second": 4.713,23      "step": 20024    },25    {26      "epoch": 1.49,27      "learning_rate": 3.8066465256797584e-05,28      "loss": 0.4748,29      "step": 40030    },31    {32      "epoch": 1.49,33      "eval_loss": 0.7374169826507568,34      "eval_matthews_correlation": 0.295093127326412,35      "eval_runtime": 4.195,36      "eval_samples_per_second": 248.628,37      "eval_steps_per_second": 7.866,38      "step": 40039    },40    {41      "epoch": 2.24,42      "learning_rate": 3.685800604229608e-05,43      "loss": 0.4165,44      "step": 60045    },46    {47      "epoch": 2.24,48      "eval_loss": 0.7154484391212463,49      "eval_matthews_correlation": 0.4656950548188858,50      "eval_runtime": 4.0204,51      "eval_samples_per_second": 259.424,52      "eval_steps_per_second": 8.208,53      "step": 60054    },55    {56      "epoch": 2.99,57      "learning_rate": 3.564954682779456e-05,58      "loss": 0.3461,59      "step": 80060    },61    {62      "epoch": 2.99,63      "eval_loss": 0.5880667567253113,64      "eval_matthews_correlation": 0.4966579188966687,65      "eval_runtime": 4.1325,66      "eval_samples_per_second": 252.389,67      "eval_steps_per_second": 7.985,68      "step": 80069    },70    {71      "epoch": 3.73,72      "learning_rate": 3.4441087613293056e-05,73      "loss": 0.2666,74      "step": 100075    },76    {77      "epoch": 3.73,78      "eval_loss": 0.6024960875511169,79      "eval_matthews_correlation": 0.5243351914860497,80      "eval_runtime": 4.2714,81      "eval_samples_per_second": 244.185,82      "eval_steps_per_second": 7.726,83      "step": 100084    },85    {86      "epoch": 4.48,87      "learning_rate": 3.323262839879154e-05,88      "loss": 0.233,89      "step": 120090    },91    {92      "epoch": 4.48,93      "eval_loss": 0.7757831811904907,94      "eval_matthews_correlation": 0.5180330680945964,95      "eval_runtime": 5.4027,96      "eval_samples_per_second": 193.052,97      "eval_steps_per_second": 6.108,98      "step": 120099    },100    {101      "epoch": 5.22,102      "learning_rate": 3.2024169184290035e-05,103      "loss": 0.2067,104      "step": 1400105    },106    {107      "epoch": 5.22,108      "eval_loss": 0.8261585831642151,109      "eval_matthews_correlation": 0.4812305179480984,110      "eval_runtime": 4.5778,111      "eval_samples_per_second": 227.84,112      "eval_steps_per_second": 7.209,113      "step": 1400114    },115    {116      "epoch": 5.97,117      "learning_rate": 3.081570996978852e-05,118      "loss": 0.174,119      "step": 1600120    },121    {122      "epoch": 5.97,123      "eval_loss": 1.0030035972595215,124      "eval_matthews_correlation": 0.4690871215881841,125      "eval_runtime": 4.1699,126      "eval_samples_per_second": 250.124,127      "eval_steps_per_second": 7.914,128      "step": 1600129    },130    {131      "epoch": 6.72,132      "learning_rate": 2.9607250755287014e-05,133      "loss": 0.1504,134      "step": 1800135    },136    {137      "epoch": 6.72,138      "eval_loss": 1.007808804512024,139      "eval_matthews_correlation": 0.505928717066083,140      "eval_runtime": 5.5718,141      "eval_samples_per_second": 187.193,142      "eval_steps_per_second": 5.923,143      "step": 1800144    },145    {146      "epoch": 7.46,147      "learning_rate": 2.83987915407855e-05,148      "loss": 0.1261,149      "step": 2000150    },151    {152      "epoch": 7.46,153      "eval_loss": 1.0862607955932617,154      "eval_matthews_correlation": 0.4929950873963667,155      "eval_runtime": 4.1938,156      "eval_samples_per_second": 248.698,157      "eval_steps_per_second": 7.869,158      "step": 2000159    },160    {161      "epoch": 8.21,162      "learning_rate": 2.719033232628399e-05,163      "loss": 0.1141,164      "step": 2200165    },166    {167      "epoch": 8.21,168      "eval_loss": 0.9651995301246643,169      "eval_matthews_correlation": 0.5632084517291658,170      "eval_runtime": 4.2997,171      "eval_samples_per_second": 242.575,172      "eval_steps_per_second": 7.675,173      "step": 2200174    }175  ],176  "max_steps": 6700,177  "num_train_epochs": 25,178  "total_flos": 5382361281024000.0,179  "trial_name": null,180  "trial_params": null181}182