all_results.json

{
    "epoch": 0.9982631930527722,
    "eval_logits/chosen": -1.5110498666763306,
    "eval_logits/rejected": -1.5196630954742432,
    "eval_logps/chosen": -1.6859560012817383,
    "eval_logps/rejected": -2.2469725608825684,
    "eval_loss": 1.24610435962677,
    "eval_rewards/accuracies": 0.8373983502388,
    "eval_rewards/chosen": -16.859560012817383,
    "eval_rewards/margins": 5.610167026519775,
    "eval_rewards/rejected": -22.4697265625,
    "eval_runtime": 87.9363,
    "eval_samples": 1961,
    "eval_samples_per_second": 22.3,
    "eval_steps_per_second": 1.399,
    "total_flos": 0.0,
    "train_loss": 1.8360214427400707,
    "train_runtime": 11486.9698,
    "train_samples": 59876,
    "train_samples_per_second": 5.213,
    "train_steps_per_second": 0.041
}