namakoo

DPO / RLHF preference datasets • code failure analysis • local LLM data factories • code validation • synthetic training data • Qwen 2.5-Coder pipelines