Skip to content

Conversation

@LeiWang1999
Copy link
Contributor

@LeiWang1999 LeiWang1999 commented Jul 1, 2024

This pull request includes a change to the install.sh script to export the TORCH_CUDA_ARCH_LIST variables in install.sh scripts. Otherwise, the build pipeline of cuda backend will crash.

Also remove importlib from requirements.txt ref to #132

@Qubitium
Copy link
Collaborator

Qubitium commented Jul 1, 2024

@LeiWang1999 Thanks! LGTM. I removed PTX flag since we don't want JIT of the cuda codes for now. We observed slightly faster code with PTX disabled.

@Qubitium Qubitium merged commit ac666be into ModelCloud:main Jul 1, 2024
DeJoker pushed a commit to DeJoker/GPTQModel that referenced this pull request Jul 19, 2024
* chore: update TORCH_CUDA_ARCH_LIST in install.sh

* chore: remove importlib from requirements.txt

* chore: Fix multi-gpu environment check in bitblas_target_detector.py

* chore: Update logger.info message in qlinear_bitblas.py

* Update install.sh

We like to not use PTX compilation for now.

---------

Co-authored-by: Qubitium-ModelCloud <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants