all_results.json

{
    "epoch": 2.999297541394882,
    "eval_logits/chosen": -0.13737329840660095,
    "eval_logits/rejected": -0.060397788882255554,
    "eval_logps/chosen": -4.343520641326904,
    "eval_logps/rejected": -5.11142110824585,
    "eval_loss": 0.05006524920463562,
    "eval_rewards/accuracies": 0.6810088753700256,
    "eval_rewards/chosen": -4.343520641326904,
    "eval_rewards/margins": 0.7678999900817871,
    "eval_rewards/rejected": -5.11142110824585,
    "eval_runtime": 44.4589,
    "eval_samples": 1345,
    "eval_samples_per_second": 30.253,
    "eval_sft_loss": 3.9426960945129395,
    "eval_steps_per_second": 7.58,
    "total_flos": 0.0,
    "train_loss": 0.0595120464655691,
    "train_runtime": 31911.443,
    "train_samples": 59790,
    "train_samples_per_second": 5.621,
    "train_steps_per_second": 0.176
}