all_results.json

{
    "epoch": 0.985781990521327,
    "eval_logits/chosen": -0.5211005210876465,
    "eval_logits/rejected": -0.7105662226676941,
    "eval_logps/chosen": -442.4708557128906,
    "eval_logps/rejected": -418.5675354003906,
    "eval_loss": 0.6833909749984741,
    "eval_rewards/accuracies": 0.59375,
    "eval_rewards/chosen": -0.008601006120443344,
    "eval_rewards/margins": 0.021825773641467094,
    "eval_rewards/rejected": -0.03042677976191044,
    "eval_runtime": 8.5409,
    "eval_samples": 750,
    "eval_samples_per_second": 87.812,
    "eval_steps_per_second": 2.81,
    "total_flos": 0.0,
    "train_loss": 0.6860739244864538,
    "train_runtime": 316.322,
    "train_samples": 6750,
    "train_samples_per_second": 21.339,
    "train_steps_per_second": 0.164
}