all_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -2.1211588382720947,
    "eval_logits/rejected": -2.000145435333252,
    "eval_logps/chosen": -265.46588134765625,
    "eval_logps/rejected": -224.6123809814453,
    "eval_loss": 0.5657259225845337,
    "eval_rewards/accuracies": 0.7020000219345093,
    "eval_rewards/chosen": -0.08337792754173279,
    "eval_rewards/margins": 0.44957080483436584,
    "eval_rewards/rejected": -0.5329487323760986,
    "eval_runtime": 600.8868,
    "eval_samples": 2000,
    "eval_samples_per_second": 3.328,
    "eval_steps_per_second": 0.208,
    "train_loss": 0.6034470564241908,
    "train_runtime": 84521.5132,
    "train_samples": 61966,
    "train_samples_per_second": 2.199,
    "train_steps_per_second": 0.009
}