all_results.json

{
    "epoch": 1.971563981042654,
    "eval_logits/chosen": 100.6234130859375,
    "eval_logits/rejected": 94.8508529663086,
    "eval_logps/chosen": -420.0129089355469,
    "eval_logps/rejected": -446.2936096191406,
    "eval_loss": 0.4608144462108612,
    "eval_rewards/accuracies": 0.7395833134651184,
    "eval_rewards/chosen": -2.8172361850738525,
    "eval_rewards/margins": 1.3938981294631958,
    "eval_rewards/rejected": -4.211134433746338,
    "eval_runtime": 121.9636,
    "eval_samples": 750,
    "eval_samples_per_second": 6.149,
    "eval_steps_per_second": 0.197,
    "total_flos": 0.0,
    "train_loss": 0.41261253735193837,
    "train_runtime": 2174.4689,
    "train_samples": 6750,
    "train_samples_per_second": 6.208,
    "train_steps_per_second": 0.048
}