all_results.json

{
    "epoch": 0.99,
    "eval_logits/chosen": 106.62171173095703,
    "eval_logits/rejected": 100.2883529663086,
    "eval_logps/chosen": -476.6144104003906,
    "eval_logps/rejected": -493.8050231933594,
    "eval_loss": 0.5058174729347229,
    "eval_rewards/accuracies": 0.7083333134651184,
    "eval_rewards/chosen": -1.1604968309402466,
    "eval_rewards/margins": 0.9522747993469238,
    "eval_rewards/rejected": -2.11277174949646,
    "eval_runtime": 51.3345,
    "eval_samples": 750,
    "eval_samples_per_second": 14.61,
    "eval_steps_per_second": 0.468,
    "train_loss": 0.5840168778712933,
    "train_runtime": 572.7421,
    "train_samples": 6750,
    "train_samples_per_second": 11.785,
    "train_steps_per_second": 0.091
}