all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.2450356483459473,
    "eval_logits/rejected": -2.2401509284973145,
    "eval_logps/chosen": -34.03712844848633,
    "eval_logps/rejected": -37.50889587402344,
    "eval_loss": 0.6972227096557617,
    "eval_rewards/accuracies": 0.4759136438369751,
    "eval_rewards/chosen": -0.0012873795349150896,
    "eval_rewards/margins": -0.005148096010088921,
    "eval_rewards/rejected": 0.0038607169408351183,
    "eval_runtime": 145.779,
    "eval_samples": 343,
    "eval_samples_per_second": 2.353,
    "eval_steps_per_second": 0.295,
    "train_loss": 0.6914243283209862,
    "train_runtime": 3256.6129,
    "train_samples": 3079,
    "train_samples_per_second": 0.945,
    "train_steps_per_second": 0.118
}