all_results.json

{
    "epoch": 0.998691442030882,
    "eval_logits/chosen": -0.7125873565673828,
    "eval_logits/rejected": -0.7502822279930115,
    "eval_logps/chosen": -405.15185546875,
    "eval_logps/rejected": -435.5634460449219,
    "eval_loss": 0.5485080480575562,
    "eval_rewards/accuracies": 0.7379999756813049,
    "eval_rewards/chosen": -1.143401861190796,
    "eval_rewards/margins": 0.5365331768989563,
    "eval_rewards/rejected": -1.679935097694397,
    "eval_runtime": 355.0474,
    "eval_samples": 2000,
    "eval_samples_per_second": 5.633,
    "eval_steps_per_second": 1.408,
    "total_flos": 0.0,
    "train_loss": 0.5814160232274037,
    "train_runtime": 29745.7246,
    "train_samples": 61134,
    "train_samples_per_second": 2.055,
    "train_steps_per_second": 0.016
}