Qwen2.5-3B-Instruct-Turkish-DPO / dpo_metadata.json
yusufbaykaloglu's picture
Upload optimized DPO model (step-400) to prevent overfitting 🚀
b8932b1 verified
{
"model_type": "DPO_ALIGNED_STEP400_FULL",
"base_model": "yusufbaykaloglu/qwen25-turkish-sft-safe",
"checkpoint_used": "step-400",
"merge_status": "FULL_MODEL_MERGED",
"expected_size": "2GB+",
"dpo_beta": 0.1
}