all_results.json

{
    "epoch": 0.995910949568378,
    "eval_logits/chosen": -6.213160991668701,
    "eval_logits/rejected": -6.001035213470459,
    "eval_logps/chosen": -0.544950544834137,
    "eval_logps/rejected": -0.6353368759155273,
    "eval_loss": 4.511509418487549,
    "eval_rewards/accuracies": 0.5963114500045776,
    "eval_rewards/chosen": -5.44950532913208,
    "eval_rewards/margins": 0.9038640260696411,
    "eval_rewards/rejected": -6.353369235992432,
    "eval_runtime": 86.5154,
    "eval_samples": 1941,
    "eval_samples_per_second": 22.435,
    "eval_steps_per_second": 1.41,
    "total_flos": 0.0,
    "train_loss": 5.239574314033898,
    "train_runtime": 2514.0727,
    "train_samples": 17605,
    "train_samples_per_second": 7.003,
    "train_steps_per_second": 0.054
}