all_results.json

{
    "epoch": 0.9982631930527722,
    "eval_logits/chosen": -1.462799310684204,
    "eval_logits/rejected": -1.471309781074524,
    "eval_logps/chosen": -1.7769458293914795,
    "eval_logps/rejected": -2.3649635314941406,
    "eval_loss": 1.2554945945739746,
    "eval_rewards/accuracies": 0.8394308686256409,
    "eval_rewards/chosen": -17.76945686340332,
    "eval_rewards/margins": 5.880179405212402,
    "eval_rewards/rejected": -23.64963722229004,
    "eval_runtime": 87.6765,
    "eval_samples": 1961,
    "eval_samples_per_second": 22.366,
    "eval_steps_per_second": 1.403,
    "total_flos": 0.0,
    "train_loss": 1.8863488443403225,
    "train_runtime": 11541.8743,
    "train_samples": 59876,
    "train_samples_per_second": 5.188,
    "train_steps_per_second": 0.04
}