all_results.json

{
    "epoch": 0.9880609304240429,
    "eval_logits/chosen": 9.158658981323242,
    "eval_logits/rejected": 9.958081245422363,
    "eval_logps/chosen": -110.41038513183594,
    "eval_logps/ref_chosen": -106.16099548339844,
    "eval_logps/ref_rejected": -119.78473663330078,
    "eval_logps/rejected": -124.09846496582031,
    "eval_loss": 0.6929683089256287,
    "eval_rewards/accuracies": 0.4972222149372101,
    "eval_rewards/chosen": -0.04249410331249237,
    "eval_rewards/margins": 0.0006433044327422976,
    "eval_rewards/rejected": -0.043137405067682266,
    "eval_runtime": 159.1867,
    "eval_samples": 1080,
    "eval_samples_per_second": 6.784,
    "eval_steps_per_second": 1.696,
    "total_flos": 0.0,
    "train_loss": 0.6922952850659688,
    "train_runtime": 7631.3864,
    "train_samples": 9715,
    "train_samples_per_second": 1.273,
    "train_steps_per_second": 0.01
}