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

{
    "epoch": 0.9999813887699838,
    "total_flos": 0.0,
    "train_loss": 0.0,
    "train_runtime": 37.9451,
    "train_samples": 3438778,
    "train_samples_per_second": 90625.095,
    "train_steps_per_second": 707.997
}