all_results.json

{
    "epoch": 2.0,
    "eval_logits/chosen": 0.1892445683479309,
    "eval_logits/rejected": 0.05091705545783043,
    "eval_logps/chosen": -458.9963073730469,
    "eval_logps/rejected": -422.9080810546875,
    "eval_loss": 0.6194455623626709,
    "eval_rewards/accuracies": 0.6458333134651184,
    "eval_rewards/chosen": -0.6699296832084656,
    "eval_rewards/margins": 0.4407287836074829,
    "eval_rewards/rejected": -1.1106584072113037,
    "eval_runtime": 150.7643,
    "eval_samples": 1143,
    "eval_samples_per_second": 7.581,
    "eval_steps_per_second": 0.239,
    "total_flos": 0.0,
    "train_loss": 0.5581450056080794,
    "train_runtime": 39294.0243,
    "train_samples_per_second": 2.619,
    "train_steps_per_second": 0.041
}