all_results.json

{
    "epoch": 2.955223880597015,
    "eval_logits/chosen": -2.3587028980255127,
    "eval_logits/rejected": -2.3534066677093506,
    "eval_logps/chosen": -267.1278381347656,
    "eval_logps/rejected": -213.8096466064453,
    "eval_loss": 0.5255125761032104,
    "eval_rewards/accuracies": 0.8583333492279053,
    "eval_rewards/chosen": 2.4023120403289795,
    "eval_rewards/margins": 2.7143805027008057,
    "eval_rewards/rejected": -0.3120681643486023,
    "eval_runtime": 61.3371,
    "eval_samples_per_second": 15.488,
    "eval_steps_per_second": 0.245,
    "total_flos": 1167029855846400.0,
    "train_loss": 0.3387684075519292,
    "train_runtime": 3422.3866,
    "train_samples_per_second": 7.492,
    "train_steps_per_second": 0.029
}