all_results.json

{
    "epoch": 2.0,
    "eval_logits/chosen": -2.3846213817596436,
    "eval_logits/rejected": -2.3414511680603027,
    "eval_logps/chosen": -80.37240600585938,
    "eval_logps/rejected": -253.9702911376953,
    "eval_loss": 0.17943651974201202,
    "eval_rewards/accuracies": 0.8947368264198303,
    "eval_rewards/chosen": 0.25255322456359863,
    "eval_rewards/margins": 7.301982879638672,
    "eval_rewards/rejected": -7.049429416656494,
    "eval_runtime": 27.9087,
    "eval_samples": 301,
    "eval_samples_per_second": 10.785,
    "eval_steps_per_second": 0.681,
    "train_loss": 0.21607095603003493,
    "train_runtime": 14362.1771,
    "train_samples": 29893,
    "train_samples_per_second": 4.163,
    "train_steps_per_second": 0.065
}