all_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -1.4300650358200073,
    "eval_logits/rejected": -1.3494709730148315,
    "eval_logps/chosen": -90.04821014404297,
    "eval_logps/rejected": -118.67648315429688,
    "eval_loss": 0.009333602152764797,
    "eval_rewards/accuracies": 0.9916201233863831,
    "eval_rewards/chosen": -7.033120155334473,
    "eval_rewards/margins": 22.35300636291504,
    "eval_rewards/rejected": -29.386123657226562,
    "eval_runtime": 171.6798,
    "eval_samples": 2862,
    "eval_samples_per_second": 16.671,
    "eval_steps_per_second": 1.043,
    "train_loss": 0.019671504644591626,
    "train_runtime": 60672.7973,
    "train_samples": 140201,
    "train_samples_per_second": 6.932,
    "train_steps_per_second": 0.108
}