Replies: 3 comments 8 replies
-
I believe Colab currently uses cuda110 wheels for jaxlib, so a first order of business before making this change would be to ensure Colab is compatible with updated versions. Edit: it appears that Colab also uses pytorch's cuda 10.2 wheels, so it may be as easy as updating the jaxlib wheels in the Colab container, and not having to touch cuda (which is a more involved change) |
Beta Was this translation helpful? Give feedback.
-
Is there a test I can run by installing the current 11.2 jaxlib in colab and making sure jax runs? AFAIK, cuda is backwards compatible for 11.x versions. |
Beta Was this translation helpful? Give feedback.
-
Now the official release for jaxlibs will be for Cuda 10.2 and Cuda 11.1 (similar to Pytorch) |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Hello everyone,
I have been thinking about releasing GPU jaxlib wheels only for cuda 10.2 and 11.2. This makes it in line with Pytorch and TF.
What does everyone think?
Beta Was this translation helpful? Give feedback.
All reactions