all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.2767388820648193,
    "eval_logits/rejected": -2.215937614440918,
    "eval_logps/chosen": -364.837646484375,
    "eval_logps/rejected": -464.28753662109375,
    "eval_loss": 0.5073517560958862,
    "eval_rewards/accuracies": 0.765625,
    "eval_rewards/chosen": -1.1322153806686401,
    "eval_rewards/margins": 0.9668433666229248,
    "eval_rewards/rejected": -2.0990586280822754,
    "eval_runtime": 95.8711,
    "eval_samples": 2000,
    "eval_samples_per_second": 20.861,
    "eval_steps_per_second": 0.334,
    "total_flos": 0.0,
    "train_loss": 0.5515718292903051,
    "train_runtime": 10135.9715,
    "train_samples": 61134,
    "train_samples_per_second": 6.031,
    "train_steps_per_second": 0.189
}