all_results.json

{
    "epoch": 4.0,
    "eval_logits/chosen": -2.22914457321167,
    "eval_logits/rejected": -2.224289894104004,
    "eval_logps/chosen": -34.094017028808594,
    "eval_logps/rejected": -37.60322189331055,
    "eval_loss": 0.6942082643508911,
    "eval_rewards/accuracies": 0.5099667906761169,
    "eval_rewards/chosen": -0.047573190182447433,
    "eval_rewards/margins": 0.021706854924559593,
    "eval_rewards/rejected": -0.06928005069494247,
    "eval_runtime": 145.4732,
    "eval_samples": 343,
    "eval_samples_per_second": 2.358,
    "eval_steps_per_second": 0.296,
    "train_loss": 0.20605388523696305,
    "train_runtime": 10804.1609,
    "train_samples": 3079,
    "train_samples_per_second": 1.14,
    "train_steps_per_second": 0.143
}