all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.2436320781707764,
    "eval_logits/rejected": -2.2387471199035645,
    "eval_logps/chosen": -34.02302932739258,
    "eval_logps/rejected": -37.51007080078125,
    "eval_loss": 0.6931225061416626,
    "eval_rewards/accuracies": 0.5328072905540466,
    "eval_rewards/chosen": 0.00011527122114785016,
    "eval_rewards/margins": 4.97526052640751e-05,
    "eval_rewards/rejected": 6.551862315973267e-05,
    "eval_runtime": 145.6938,
    "eval_samples": 343,
    "eval_samples_per_second": 2.354,
    "eval_steps_per_second": 0.295,
    "train_loss": 0.6930816006350827,
    "train_runtime": 3252.6347,
    "train_samples": 3079,
    "train_samples_per_second": 0.947,
    "train_steps_per_second": 0.118
}