qwen2.5-3b-dpo-finegrained/train_results.json

{
    "epoch": 1.0,
    "total_flos": 0.0,
    "train_loss": 0.0,
    "train_runtime": 39.4392,
    "train_samples": 1773152,
    "train_samples_per_second": 44959.109,
    "train_steps_per_second": 351.249
}