zephyr-7b-dpo-qlora / all_results.json
objects76's picture
End of training
e0d3a4a verified
raw
history blame contribute delete
741 Bytes
{
"epoch": 1.0,
"eval_logits/chosen": -1.2524774074554443,
"eval_logits/rejected": -1.1317628622055054,
"eval_logps/chosen": -545.8759765625,
"eval_logps/rejected": -635.6185302734375,
"eval_loss": 0.4863443076610565,
"eval_rewards/accuracies": 0.7394999861717224,
"eval_rewards/chosen": -2.812239646911621,
"eval_rewards/margins": 1.0978591442108154,
"eval_rewards/rejected": -3.9100985527038574,
"eval_runtime": 384.8212,
"eval_samples": 2000,
"eval_samples_per_second": 5.197,
"eval_steps_per_second": 0.65,
"train_loss": 0.5238101308459981,
"train_runtime": 42749.2467,
"train_samples": 61134,
"train_samples_per_second": 1.43,
"train_steps_per_second": 0.089
}