倉庫指標
- 星標
- (1,244 顆星)
- PR 合併指標
- (PR 指標待抓取)
描述
‼️ If you want to work on this issue: please comment below and wait until a maintainer assigns this issue to you before opening a PR to avoid several contributions on the same issue. Thanks! 😊
✨ What You’ll Do
Making the models faster, smaller, and greener is essential in Pruna, which is why we continually seek new ways to increase efficiency.
LLM Compresor is a library for optimizing models with vllm. Right now, we have only integrated AWQ, but it would be nice to implement the rest of quantizers: Simple PTQ, GPTQ, SmoothQuant, or SparseGPT.
Don't hesitate to coordinate and collaborate to work on each of them.
🤖 Useful Resources
- Don't hesitate to consult our guide on how to implement an algorithm: https://docs.pruna.ai/en/stable/docs_pruna/user_manual/customize_algorithm.html#step-4-add-algorithm-attributes
- https://github.com/vllm-project/llm-compressor
- Explore the codebase and get familiar with the main concepts
- Read the guide and start implementing the new algo
- Test and write the documentation
✅ Acceptance Criteria
{Which are the requirements for the acceptance? Should the code work in any specific way? What’s the correct output}
- Tests & Docs: All existing and new unit tests pass, and the documentation is updated
And don’t forget to give us a ⭐️!
❓ Questions?
Feel free to jump into the #contributing Discord channel] if you hit any roadblocks. Can’t wait to see your contribution! 🚀
