all_results.json

{
    "epoch": 2.97,
    "eval_logits/chosen": -2.256798028945923,
    "eval_logits/rejected": -2.121875524520874,
    "eval_logps/chosen": -408.6165466308594,
    "eval_logps/rejected": -364.0181884765625,
    "eval_loss": 0.6663674712181091,
    "eval_rewards/accuracies": 0.6666666865348816,
    "eval_rewards/chosen": -0.06976744532585144,
    "eval_rewards/margins": 0.05996997281908989,
    "eval_rewards/rejected": -0.12973742187023163,
    "eval_runtime": 39.1746,
    "eval_samples": 201,
    "eval_samples_per_second": 5.131,
    "eval_steps_per_second": 1.71,
    "train_loss": 0.6770667407682033,
    "train_runtime": 12894.8655,
    "train_samples": 9845,
    "train_samples_per_second": 2.29,
    "train_steps_per_second": 0.014
}