Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Separate AutoGPTQ dep to pip install -e .[auto-gptq] #1077

Merged
merged 2 commits into from
Jan 9, 2024

Conversation

casper-hansen
Copy link
Collaborator

This fixes the following issue:

  1. You already have the correct version of torch installed with CUDA dependencies and everything.
  2. You install axolotl with pip install -e .
  3. The AutoGPTQ installation proceeds to
    • redownload torch
    • redownload CUDA dependencies

This issue causes bloating of the environment and longer build times. So, I have separated this dependency as it's frankly not needed for 99% of training runs.

setup.py Outdated
@@ -61,5 +61,8 @@ def parse_requirements():
"mamba-ssm": [
"mamba-ssm==1.0.1",
],
"autogptq": [
"auto-gptq==0.5.1 --extra-index-url https://huggingface.github.io/autogptq-index/whl/cu118/",
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

keep --extra-index-url https://huggingface.github.io/autogptq-index/whl/cu118/ in the requirements.txt, and this line is simply

Suggested change
"auto-gptq==0.5.1 --extra-index-url https://huggingface.github.io/autogptq-index/whl/cu118/",
"auto-gptq==0.5.1",

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Fixed it and also fixed it so that it's pip install -e .[auto-gptq] as the original package is named auto-gptq.

@casper-hansen casper-hansen changed the title Separate AutoGPTQ dep to pip install -e .[autogptq] Separate AutoGPTQ dep to pip install -e .[auto-gptq] Jan 9, 2024
@winglian
Copy link
Collaborator

winglian commented Jan 9, 2024

btw, I think the docker tests are failing mostly b/c I fixed the runner, so now it properly exposes CUDA

@casper-hansen casper-hansen merged commit 9be92d1 into main Jan 9, 2024
4 of 6 checks passed
@winglian winglian deleted the separate_autogptq branch January 23, 2024 12:33
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.

None yet

2 participants