Run 35B LLMs on Dual Pascal GPUs with QLoRA | Not Hacker News!