all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -0.9538615942001343,
    "eval_logits/rejected": -0.9195562601089478,
    "eval_logps/chosen": -369.7508544921875,
    "eval_logps/rejected": -379.8497314453125,
    "eval_loss": 0.03875432908535004,
    "eval_rewards/accuracies": 0.6069999933242798,
    "eval_rewards/chosen": 0.02655443549156189,
    "eval_rewards/margins": 0.03916873782873154,
    "eval_rewards/rejected": -0.012614301405847073,
    "eval_runtime": 539.9666,
    "eval_samples": 2000,
    "eval_samples_per_second": 3.704,
    "eval_steps_per_second": 0.926,
    "train_loss": 0.08419492204701015,
    "train_runtime": 22113.8812,
    "train_samples": 61135,
    "train_samples_per_second": 1.131,
    "train_steps_per_second": 0.071
}