all_results.json

{
    "epoch": 1.971563981042654,
    "eval_logits/chosen": 113.6686782836914,
    "eval_logits/rejected": 108.56292724609375,
    "eval_logps/chosen": -332.4104919433594,
    "eval_logps/rejected": -363.9787902832031,
    "eval_loss": 0.29782694578170776,
    "eval_rewards/accuracies": 0.6979166865348816,
    "eval_rewards/chosen": 1.5628873109817505,
    "eval_rewards/margins": 1.6582797765731812,
    "eval_rewards/rejected": -0.09539230912923813,
    "eval_runtime": 121.439,
    "eval_samples": 750,
    "eval_samples_per_second": 6.176,
    "eval_steps_per_second": 0.198,
    "total_flos": 0.0,
    "train_loss": 0.29816230271871275,
    "train_runtime": 2165.9381,
    "train_samples": 6750,
    "train_samples_per_second": 6.233,
    "train_steps_per_second": 0.048
}