all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -1.352773666381836,
    "eval_logits/rejected": -1.2867770195007324,
    "eval_logps/chosen": -451.33642578125,
    "eval_logps/rejected": -657.6813354492188,
    "eval_loss": 0.2841675579547882,
    "eval_rewards/accuracies": 0.88671875,
    "eval_rewards/chosen": -1.1548646688461304,
    "eval_rewards/margins": 2.4814703464508057,
    "eval_rewards/rejected": -3.6363346576690674,
    "eval_runtime": 527.3224,
    "eval_samples": 1536,
    "eval_samples_per_second": 2.913,
    "eval_steps_per_second": 0.728,
    "train_loss": 0.3634731802597361,
    "train_runtime": 16591.2723,
    "train_samples": 18653,
    "train_samples_per_second": 1.124,
    "train_steps_per_second": 0.07
}