all_results.json

{
    "epoch": 0.9982631930527722,
    "eval_logits/chosen": -1.3886725902557373,
    "eval_logits/rejected": -1.3989359140396118,
    "eval_logps/chosen": -1.5420804023742676,
    "eval_logps/rejected": -2.069298505783081,
    "eval_loss": 1.2707444429397583,
    "eval_rewards/accuracies": 0.8373983502388,
    "eval_rewards/chosen": -15.42080307006836,
    "eval_rewards/margins": 5.272182941436768,
    "eval_rewards/rejected": -20.69298553466797,
    "eval_runtime": 119.8928,
    "eval_samples": 1961,
    "eval_samples_per_second": 16.356,
    "eval_steps_per_second": 1.026,
    "total_flos": 0.0,
    "train_loss": 1.868879672036181,
    "train_runtime": 17907.3194,
    "train_samples": 59876,
    "train_samples_per_second": 3.344,
    "train_steps_per_second": 0.026
}