all_results.json

{
    "epoch": 4.0,
    "eval_logits/chosen": -2.205192804336548,
    "eval_logits/rejected": -2.20035719871521,
    "eval_logps/chosen": -34.1010856628418,
    "eval_logps/rejected": -37.70829772949219,
    "eval_loss": 6.077866077423096,
    "eval_rewards/accuracies": 0.5714285373687744,
    "eval_rewards/chosen": -0.013306654058396816,
    "eval_rewards/margins": 0.025028489530086517,
    "eval_rewards/rejected": -0.03833514451980591,
    "eval_runtime": 145.6887,
    "eval_samples": 343,
    "eval_samples_per_second": 2.354,
    "eval_steps_per_second": 0.295,
    "train_loss": 1.3887401007986688,
    "train_runtime": 10806.9485,
    "train_samples": 3079,
    "train_samples_per_second": 1.14,
    "train_steps_per_second": 0.143
}