all_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -2.2166244983673096,
    "eval_logits/rejected": -2.176083564758301,
    "eval_logps/chosen": -260.0429382324219,
    "eval_logps/rejected": -234.69407653808594,
    "eval_loss": 0.5717999339103699,
    "eval_rewards/accuracies": 0.7875000238418579,
    "eval_rewards/chosen": 1.6485886573791504,
    "eval_rewards/margins": 2.3444631099700928,
    "eval_rewards/rejected": -0.6958744525909424,
    "eval_runtime": 127.4684,
    "eval_samples_per_second": 14.906,
    "eval_steps_per_second": 0.235,
    "total_flos": 2369906314051584.0,
    "train_loss": 0.3513164082866403,
    "train_runtime": 7398.18,
    "train_samples_per_second": 6.932,
    "train_steps_per_second": 0.027
}