all_results.json

{
    "epoch": 6.0,
    "eval_logits/chosen": 3.4052021503448486,
    "eval_logits/rejected": 3.43179988861084,
    "eval_logps/chosen": -634.6866455078125,
    "eval_logps/rejected": -618.552978515625,
    "eval_loss": 0.4747713804244995,
    "eval_rewards/accuracies": 0.7936508059501648,
    "eval_rewards/chosen": -0.913497269153595,
    "eval_rewards/margins": 1.0312875509262085,
    "eval_rewards/rejected": -1.9447849988937378,
    "eval_runtime": 203.6628,
    "eval_samples": 2000,
    "eval_samples_per_second": 9.82,
    "eval_steps_per_second": 0.309,
    "train_loss": 0.4762616708060343,
    "train_runtime": 91379.3933,
    "train_samples": 58917,
    "train_samples_per_second": 3.869,
    "train_steps_per_second": 0.06
}