all_results.json

{
    "epoch": 2.0,
    "eval_logits/chosen": -3.11618709564209,
    "eval_logits/rejected": -3.1104962825775146,
    "eval_logps/chosen": -59.58960723876953,
    "eval_logps/rejected": -64.70088958740234,
    "eval_loss": 0.6901275515556335,
    "eval_rewards/accuracies": 0.589219331741333,
    "eval_rewards/chosen": -0.00877712108194828,
    "eval_rewards/margins": 0.006430591456592083,
    "eval_rewards/rejected": -0.015207710675895214,
    "eval_runtime": 383.5579,
    "eval_samples": 4304,
    "eval_samples_per_second": 11.221,
    "eval_steps_per_second": 1.403,
    "total_flos": 0.0,
    "train_loss": 0.6883086910911629,
    "train_runtime": 95005.2512,
    "train_samples": 92858,
    "train_samples_per_second": 1.955,
    "train_steps_per_second": 0.122
}