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

add 'conda install cudatoolkit' to dreambooth 'training on 16GB' example #1229

Merged
merged 2 commits into from
Nov 15, 2022

Conversation

0xdevalias
Copy link
Contributor

fixes #1207

@HuggingFaceDocBuilderDev

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint.

@patrickvonplaten
Copy link
Contributor

Hmm,

I'm always a bit worried about giving clear instructions how to install CUDA as everybody does it differently and I think we should just leave it as "use bitsandbytes with GPU" and then it's a bit up to bitsandbytes to make sure everything works as expected no?

@patil-suraj @pcuenca what do you think?

@0xdevalias
Copy link
Contributor Author

0xdevalias commented Nov 9, 2022

I don't really mind if this doesn't end up being the exact solution, but currently the example docs say "do this and it will work", it doesn't work, and I spent days having to trace and track down exactly why that was.. so.. not ideal.


There is also this issue over at bitsandbytes, though I don't know that it's strictly related as most of the issues in it seem to be more "I installed this wrong":


The docs for installation of bitsandbytes explicitly mentions it as a requirement:

Requirements: anaconda, cudatoolkit, pytorch

The requirements can best be fulfilled by installing pytorch via anaconda. You can install PyTorch by following the "Get Started" instructions on the official website.


Also, when I tried to look up how to install cudatoolkit, conda is mentioned as an explicit method:

I guess technically it can be done with pip too, but that seems more convoluted:

@patil-suraj
Copy link
Contributor

Agree with @patrickvonplaten , we can't assume that everyone is using conda, I think here we could just point to bitsandbytes docs. @0xdevalias would be awesome if you could update the PR to instead add the link to bitsandbytes

@0xdevalias
Copy link
Contributor Author

would be awesome if you could update the PR to instead add the link to bitsandbytes

@patil-suraj This PR is set to allow edits by maintainers. I think it might be best if you/one of the others updated the text as you like it to be, rather than having to back and forth to get it right. Alternatively, if you make a change suggestion through GitHub's PR review, I'm happy to click apply on it.

@patil-suraj
Copy link
Contributor

Thanks @0xdevalias , just updated the instruction.

@patil-suraj patil-suraj merged commit db1cb0b into huggingface:main Nov 15, 2022
@HuggingFaceDocBuilderDev

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint.

@0xdevalias 0xdevalias deleted the patch-1 branch November 15, 2022 12:36
allo- pushed a commit to allo-/diffusers that referenced this pull request Nov 15, 2022
…e#1229)

* add 'conda install cudatoolkit' to dreambooth 'training on 16GB' example 

fixes huggingface#1207

* Apply suggestions from code review

Co-authored-by: Suraj Patil <surajp815@gmail.com>
@StatsGary
Copy link

Yes this is still affecting me and links to the bitsandbytes setup. I have tried to resolve the issue, but I am still getting errors.

I have raised an issue on the bitsandbytes site:

RuntimeError: CUDA Setup failed despite GPU being available. Inspect the CUDA SETUP outputs aboveto fix your environment! If you cannot find any issues and suspect a bug, please open an issue with detals about your environment: https://github.com/TimDettmers/bitsandbytes/issues

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.

AttributeError: /opt/conda/bin/python: undefined symbol: cudaRuntimeGetVersion
5 participants