-
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
jax[cuda] installation replaces current jax version with old jax-0.2.22 version #12307
Comments
|
Actually I'm wrong! You had the right URL the first time. I think the index was in a broken state. Try now? @yashk2810 Is it possible the index file doesn't get updated atomically? |
@hawkinsp still getting the same output! |
My guess is that's related to some sort of caching of the index and it will fix itself soon. If you open it up in a web browser, do you see:
in the list? You can always download the necessary wheel manually. |
Yes, they are all there. |
Oh! Wait! You are on a Mac. We don't support CUDA on Mac. So there's no matching wheel found. You should install the CPU wheels on Mac (i.e., just |
I'm seeing the same issue on Ubuntu as well. Most recently is the behavior below on a new Jetson Orin Dev Kit (so it is Linux for Tortuga)
But if I instead provide my versions of cuda (11.4) and cuddnn (8.3.2). I get the expected behavior:
If I just just
Is this just that |
@treyra What version of Python are you using? |
Also, what happens if you run this?
|
Thanks for replying so fast! I'm using python 3.8.10. (And pip is 22.3) Running that command gives me:
|
That wheel is in the index and the command works on Colab. So there must be some quirk in your own system. A couple guesses:
|
Thank you, that might be the issue. When I try to just pip install jax, it does install jax 0.3.23, but not jaxlib along side it. I looks like from #7097 jax isn't compatible with ARM CPUs? The Nvidia Jetson Orin uses an Arm Cortex CPU. I'm guessing that is why above it says you don't support mac? Maybe this should be a separate issue or mentioned on #7097, but I didn't see that jax wheels are only available of x86_64 architectures until digging into the issues, is that something that makes sense to add to the installation guide? |
Yes, if the steps here didn't help with your problem, we should add a warning there about the issue you ran into. I'm not sure about the status of jaxlib wheel releases with ARM CPUs, but it sound like that may be the culprit. |
Thank you for your help with this, I think that is the main issue. I found #7186 about compiling jax for an older version of the Jetson series, so I'll try following that. The steps you linked were what I was following when I ran into this, I think adding a warning about ARM CPUs would fix the confusion. I can try submitting a pull request on that if you would like. |
Yes, a PR would be much appreciated. Thanks! |
Currently non x86_64 linux architectures are not supported, see jax-ml#7097 for request to change this. This can lead to installation confusion, as jax will install, but jaxlib will not. For example see jax-ml#12307. If this can be more clearly phrased or explained, let me know.
Currently non x86_64 linux architectures are not supported, see jax-ml#7097 for request to change this. This can lead to installation confusion, as jax will install, but jaxlib will not. For example see jax-ml#12307. If this can be more clearly phrased or explained, let me know.
Currently non x86_64 linux architectures are not supported, see jax-ml#7097 for request to change this. This can lead to installation confusion, as jax will install, but jaxlib will not. For example see jax-ml#12307. This adds a note to the install sections for the relevant pip wheels.
Description
When I run
pip install --upgrade "jax[cuda]" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html
on my laptop, the latest version of jax, currently jax-0.3.17, gets replaced by jax-0.2.22. I don't think this was happening before. Here is a representative output of the command above:
What jax/jaxlib version are you using?
jax-0.3.17
Which accelerator(s) are you using?
GPU
Additional System Info
MAC
The text was updated successfully, but these errors were encountered: