all_results.json

{
    "epoch": 0.9982631930527722,
    "eval_logits/chosen": -1.444885492324829,
    "eval_logits/rejected": -1.4548581838607788,
    "eval_logps/chosen": -1.649869441986084,
    "eval_logps/rejected": -2.216320276260376,
    "eval_loss": 1.3464635610580444,
    "eval_rewards/accuracies": 0.8414633870124817,
    "eval_rewards/chosen": -16.498695373535156,
    "eval_rewards/margins": 5.664506435394287,
    "eval_rewards/rejected": -22.1632022857666,
    "eval_runtime": 87.5514,
    "eval_samples": 1961,
    "eval_samples_per_second": 22.398,
    "eval_steps_per_second": 1.405,
    "total_flos": 0.0,
    "train_loss": 1.7901343150618756,
    "train_runtime": 11462.4072,
    "train_samples": 59876,
    "train_samples_per_second": 5.224,
    "train_steps_per_second": 0.041
}