all_results.json

{
    "epoch": 0.9880609304240429,
    "eval_logits/chosen": 9.331795692443848,
    "eval_logits/rejected": 10.156203269958496,
    "eval_logps/chosen": -80.09001922607422,
    "eval_logps/ref_chosen": -106.16099548339844,
    "eval_logps/ref_rejected": -119.78473663330078,
    "eval_logps/rejected": -93.58577728271484,
    "eval_loss": 0.44601914286613464,
    "eval_rewards/accuracies": 0.5018518567085266,
    "eval_rewards/chosen": 0.2607097327709198,
    "eval_rewards/margins": -0.001279854099266231,
    "eval_rewards/rejected": 0.261989563703537,
    "eval_runtime": 192.9589,
    "eval_samples": 1080,
    "eval_samples_per_second": 5.597,
    "eval_sft_loss": 0.27996528148651123,
    "eval_steps_per_second": 1.399,
    "total_flos": 0.0,
    "train_loss": 0.46543833255767825,
    "train_runtime": 8158.3981,
    "train_samples": 9715,
    "train_samples_per_second": 1.191,
    "train_steps_per_second": 0.009
}