-
Notifications
You must be signed in to change notification settings - Fork 2.8k
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
Release for CUDA 11.2 #5668
Comments
I am trying to get it working with cuda11.2 as well. I tried to build from source, but I could not manage to do it. OS: Ubuntu 20.04.1
cudnn version: 8.1.0.77 python version: 3.8.5 After following the instructions for building from source, I got the following error.
|
We can add CUDA 11.2 to our next wheel release, which will probably be in the next couple of weeks. @fcossio That looks like your copy of gcc or clang crashed. Try a different (newer?) version. |
The just-released jaxlib 0.1.61 now includes CUDA 11.2 wheels! |
It says
|
I suspect you are using a non-supported Python version such as 3.10 (see issue #8097) or a non-supported GPU architecture (see #2012), or a non-supported operating system such as Windows (#438) or MacOS ARM (#5501). You can see exactly which wheels are currently available at this link: https://storage.googleapis.com/jax-releases/jax_releases.html |
I manually downloaded and installed it. However, I think the following command also works:
|
Great! I'm not sure what might have gone wrong initially (the command from #5668 (comment) should work on supported systems) but I'm glad you found an approach that worked for you. |
@fcossio did you attempt to update your |
This is a very old issue for me and I don't remember well. I think I solved this issue by expanding the swap memory and then restarting the computer after the build was successful. Sorry that I have no more details. |
The available CUDA version I have is 11.2 and I would like to install jax on it.
Currently there is now release for jaxlib-0.1.60+cuda112.
Is it possible to build a release that supports CUDA 11.2?
Is there another way to install jax with CUDA 11.2 other than running
pip install --upgrade jax jaxlib==0.1.60+cuda112 -f https://storage.googleapis.com/jax-releases/jax_releases.html
Thanks!
The text was updated successfully, but these errors were encountered: