Post by @changliuchuan • Hey

Try to optimize the GPT first and do not perform first, so that most unnecessary waste can be saved