all_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -5.246756553649902,
    "eval_logits/rejected": -4.960155010223389,
    "eval_logps/chosen": -588.7877197265625,
    "eval_logps/rejected": -450.20184326171875,
    "eval_loss": 0.5464638471603394,
    "eval_rewards/accuracies": 0.7170000076293945,
    "eval_rewards/chosen": -0.05181777477264404,
    "eval_rewards/margins": 0.7143149375915527,
    "eval_rewards/rejected": -0.7661327123641968,
    "eval_runtime": 103.5267,
    "eval_samples": 2000,
    "eval_samples_per_second": 19.319,
    "eval_steps_per_second": 1.207,
    "train_loss": 0.49318724932753905,
    "train_runtime": 14700.0435,
    "train_samples": 61966,
    "train_samples_per_second": 12.646,
    "train_steps_per_second": 0.79
}