all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.223938465118408,
    "eval_logits/rejected": -2.2191123962402344,
    "eval_logps/chosen": -34.10564041137695,
    "eval_logps/rejected": -37.63585662841797,
    "eval_loss": 0.9714530110359192,
    "eval_rewards/accuracies": 0.5365448594093323,
    "eval_rewards/chosen": -0.042656492441892624,
    "eval_rewards/margins": 0.028885547071695328,
    "eval_rewards/rejected": -0.07154203206300735,
    "eval_runtime": 145.6066,
    "eval_samples": 343,
    "eval_samples_per_second": 2.356,
    "eval_steps_per_second": 0.295,
    "train_loss": 0.859264765157328,
    "train_runtime": 3251.7628,
    "train_samples": 3079,
    "train_samples_per_second": 0.947,
    "train_steps_per_second": 0.118
}