all_results.json

{
    "epoch": 2.97196261682243,
    "eval_logits/chosen": -2.499199151992798,
    "eval_logits/rejected": -2.470691680908203,
    "eval_logps/chosen": -247.51588439941406,
    "eval_logps/rejected": -236.08447265625,
    "eval_loss": 0.5324171185493469,
    "eval_rewards/accuracies": 0.8072916865348816,
    "eval_rewards/chosen": 1.255717396736145,
    "eval_rewards/margins": 2.643660545349121,
    "eval_rewards/rejected": -1.3879432678222656,
    "eval_runtime": 100.1437,
    "eval_samples_per_second": 15.178,
    "eval_steps_per_second": 0.24,
    "total_flos": 1874604243025920.0,
    "train_loss": 0.35437198479970294,
    "train_runtime": 5742.7927,
    "train_samples_per_second": 7.144,
    "train_steps_per_second": 0.028
}