feat(server): Add exllama GPTQ CUDA kernel support #553

[Narsil]

Just trying to get the integration tests to pass.

What does this PR do?

Fixes # (issue)

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a Github issue or the forum? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

[Atry]

Shall we also update Makefile?

[aoyifei]

I just clone the latest commit, and then install from local, how could i use these exllama for gptq? or I just launch the server with --quantize gptq, then it wiil automatically using this feature?

[Atry]

You need to use the docker image to use exllama. TGI has not yet included any build script to install exllama for development purpose.

[OlivierDehaene]

cd server/exllama_kernels
TORCH_CUDA_ARCH_LIST="8.0;8.6+PTX" python setup.py install

[BEpresent]

You need to use the docker image to use exllama. TGI have not yet included any build script to install exllama for development purpose.

Thanks, just to confirm, when I use the docker image for inference then exllama will automatically be used if it is possible with the model and with just the --quantize gptq flag ?