all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.243980884552002,
    "eval_logits/rejected": -2.2390992641448975,
    "eval_logps/chosen": -34.03550720214844,
    "eval_logps/rejected": -37.50373458862305,
    "eval_loss": 1.0085219144821167,
    "eval_rewards/accuracies": 0.48172760009765625,
    "eval_rewards/chosen": -0.0005728387623094022,
    "eval_rewards/margins": -0.008305290713906288,
    "eval_rewards/rejected": 0.007732452359050512,
    "eval_runtime": 145.8784,
    "eval_samples": 343,
    "eval_samples_per_second": 2.351,
    "eval_steps_per_second": 0.295,
    "train_loss": 0.9932638019710393,
    "train_runtime": 3254.1177,
    "train_samples": 3079,
    "train_samples_per_second": 0.946,
    "train_steps_per_second": 0.118
}