all_results.json

{
    "epoch": 2.998430141287284,
    "eval_logits/chosen": -6.25,
    "eval_logits/rejected": -4.96875,
    "eval_logps/chosen": -1004.0,
    "eval_logps/rejected": -1016.0,
    "eval_loss": 1.2351406812667847,
    "eval_rewards/accuracies": 0.537109375,
    "eval_rewards/chosen": -6.84375,
    "eval_rewards/margins": 0.404296875,
    "eval_rewards/rejected": -7.25,
    "eval_runtime": 46.8468,
    "eval_samples": 2000,
    "eval_samples_per_second": 42.692,
    "eval_steps_per_second": 0.683,
    "total_flos": 0.0,
    "train_loss": 0.2668540406601591,
    "train_runtime": 12367.4943,
    "train_samples": 61122,
    "train_samples_per_second": 14.826,
    "train_steps_per_second": 0.232
}