all_results.json

{
    "epoch": 0.9982631930527722,
    "eval_logits/chosen": -1.4905521869659424,
    "eval_logits/rejected": -1.501386284828186,
    "eval_logps/chosen": -1.7485934495925903,
    "eval_logps/rejected": -2.3540291786193848,
    "eval_loss": 1.1744760274887085,
    "eval_rewards/accuracies": 0.8434959053993225,
    "eval_rewards/chosen": -17.485933303833008,
    "eval_rewards/margins": 6.054358005523682,
    "eval_rewards/rejected": -23.54029083251953,
    "eval_runtime": 63.2847,
    "eval_samples": 1961,
    "eval_samples_per_second": 30.987,
    "eval_steps_per_second": 1.944,
    "total_flos": 0.0,
    "train_loss": 1.7882541670789045,
    "train_runtime": 8154.3088,
    "train_samples": 59876,
    "train_samples_per_second": 7.343,
    "train_steps_per_second": 0.057
}