all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.6714468002319336,
    "eval_logits/rejected": -2.631878614425659,
    "eval_logps/chosen": -350.4954528808594,
    "eval_logps/rejected": -443.81005859375,
    "eval_loss": 0.5209673643112183,
    "eval_rewards/accuracies": 0.78515625,
    "eval_rewards/chosen": -0.9161596894264221,
    "eval_rewards/margins": 0.890291690826416,
    "eval_rewards/rejected": -1.8064512014389038,
    "eval_runtime": 94.1916,
    "eval_samples": 2000,
    "eval_samples_per_second": 21.233,
    "eval_steps_per_second": 0.34,
    "total_flos": 0.0,
    "train_loss": 0.5618298566509827,
    "train_runtime": 11326.5504,
    "train_samples": 61134,
    "train_samples_per_second": 5.397,
    "train_steps_per_second": 0.169
}