You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hello, congratulations on the initiative to share the project on GitHub.
I would like to suggest that the hardware requirements be explicitly stated in the documentation.
manylinux20yy = yy year year
_x86_64.whl = X86 or amd64 architecture
In other words, little compatibility with Unix or Windows.
I'm not saying it's not possible to emulate via WSL or even containers or virtual machines, but it's not written anywhere that the support is only for Linux when using the CUDA modules.
The text was updated successfully, but these errors were encountered:
GraoMelo
added a commit
to GraoMelo/grok-1
that referenced
this issue
May 2, 2024
Hello, congratulations on the initiative to share the project on GitHub.
I would like to suggest that the hardware requirements be explicitly stated in the documentation.
When analyzing the dependencies:
jax[cuda12-pip]==0.4.25 -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html
In my view, it makes sense, but on the URL:
https://storage.googleapis.com/jax-releases/jax_cuda_releases.html
there are many files:
cuda100/jaxlib-0.1.52+cuda100-cp36-none-manylinux2010_x86_64.whl
cuda101/jaxlib-0.1.52+cuda101-cp36-none-manylinux2010_x86_64.whl
cuda102/jaxlib-0.1.52+cuda102-cp36-none-manylinux2010_x86_64.whl
cuda11/jaxlib-0.1.73+cuda11.cudnn805-cp37-none-manylinux2010_x86_64.whl
or even:
cuda110/jaxlib-0.1.52+cuda110-cp36-none-manylinux2010_x86_64.whl
cuda111/jaxlib-0.1.56+cuda111-cp36-none-manylinux2010_x86_64.whl
or still
cuda112/jaxlib-0.1.65+cuda112-cp39-none-manylinux2010_x86_64.whl
cuda12/jaxlib-0.4.10+cuda12.cudnn88-cp310-cp310-manylinux2014_x86_64.whl
All the files have the suffix:
manylinux20yy = yy year year
_x86_64.whl = X86 or amd64 architecture
In other words, little compatibility with Unix or Windows.
I'm not saying it's not possible to emulate via WSL or even containers or virtual machines, but it's not written anywhere that the support is only for Linux when using the CUDA modules.
The text was updated successfully, but these errors were encountered: