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

{
    "epoch": 1.0,
    "total_flos": 0.0,
    "train_loss": 0.0,
    "train_runtime": 43.6247,
    "train_samples": 1701200,
    "train_samples_per_second": 38996.291,
    "train_steps_per_second": 304.667
}