all_results.json

{
    "epoch": 1.98,
    "eval_logits/chosen": 0.8359472155570984,
    "eval_logits/rejected": 0.893373429775238,
    "eval_logps/chosen": -256.5486145019531,
    "eval_logps/rejected": -233.4872283935547,
    "eval_loss": 0.010263882577419281,
    "eval_rewards/accuracies": 0.5034999847412109,
    "eval_rewards/chosen": 0.00065155781339854,
    "eval_rewards/margins": 0.00013779415166936815,
    "eval_rewards/rejected": 0.0005137636326253414,
    "eval_runtime": 535.2197,
    "eval_samples": 2000,
    "eval_samples_per_second": 3.737,
    "eval_steps_per_second": 0.934,
    "train_loss": 0.01022457688926689,
    "train_runtime": 1632.1621,
    "train_samples": 61135,
    "train_samples_per_second": 1.225,
    "train_steps_per_second": 0.076
}