all_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -2.627108335494995,
    "eval_logits/rejected": -2.7526450157165527,
    "eval_logps/chosen": -280.3030700683594,
    "eval_logps/rejected": -327.84844970703125,
    "eval_loss": 0.1772317737340927,
    "eval_rewards/accuracies": 0.8384615182876587,
    "eval_rewards/chosen": -0.9390209317207336,
    "eval_rewards/margins": 3.175048828125,
    "eval_rewards/rejected": -4.114069938659668,
    "eval_runtime": 101.4123,
    "eval_samples": 258,
    "eval_samples_per_second": 2.544,
    "eval_steps_per_second": 0.641,
    "train_loss": 0.2459611980317006,
    "train_runtime": 34910.6145,
    "train_samples": 12601,
    "train_samples_per_second": 1.083,
    "train_steps_per_second": 0.135
}