all_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -2.0658671855926514,
    "eval_logits/rejected": -1.9411793947219849,
    "eval_logps/chosen": -266.13037109375,
    "eval_logps/rejected": -228.2694091796875,
    "eval_loss": 0.5265706777572632,
    "eval_rewards/accuracies": 0.7459999918937683,
    "eval_rewards/chosen": -0.1469534933567047,
    "eval_rewards/margins": 0.7515553832054138,
    "eval_rewards/rejected": -0.8985088467597961,
    "eval_runtime": 453.4186,
    "eval_samples": 2000,
    "eval_samples_per_second": 4.411,
    "eval_steps_per_second": 0.276,
    "train_loss": 0.5642705829018732,
    "train_runtime": 71197.0849,
    "train_samples": 61966,
    "train_samples_per_second": 2.611,
    "train_steps_per_second": 0.041
}