eval_results.json

{
    "epoch": 0.998691442030882,
    "eval_logits/chosen": 0.14297574758529663,
    "eval_logits/rejected": 0.009554736316204071,
    "eval_logps/chosen": -334.2958984375,
    "eval_logps/rejected": -339.11639404296875,
    "eval_loss": 0.5491579174995422,
    "eval_rewards/accuracies": 0.7120000123977661,
    "eval_rewards/chosen": -0.3376711905002594,
    "eval_rewards/margins": 0.3970130383968353,
    "eval_rewards/rejected": -0.7346842288970947,
    "eval_runtime": 86.7709,
    "eval_samples": 2000,
    "eval_samples_per_second": 23.049,
    "eval_steps_per_second": 1.441
}