all_results.json

{
    "epoch": 2.0,
    "eval_logits/chosen": 0.7796330451965332,
    "eval_logits/rejected": 0.8823836445808411,
    "eval_logps/chosen": -261.0818176269531,
    "eval_logps/rejected": -233.2046661376953,
    "eval_loss": 0.0004278263368178159,
    "eval_rewards/accuracies": 0.6700000166893005,
    "eval_rewards/chosen": -0.008376287296414375,
    "eval_rewards/margins": 0.009309147484600544,
    "eval_rewards/rejected": -0.017685433849692345,
    "eval_runtime": 323.1582,
    "eval_samples": 2000,
    "eval_samples_per_second": 6.189,
    "eval_steps_per_second": 0.387,
    "train_loss": 2.2877212975025803e-05,
    "train_runtime": 855.8096,
    "train_samples": 30567,
    "train_samples_per_second": 71.434,
    "train_steps_per_second": 1.115
}