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
{{ message }}
This repository has been archived by the owner on Aug 27, 2024. It is now read-only.
I talked with @csadorf today and it seems like there might be a useful layer between mambaforge-cuda and ci-imgs that could include the CUDA Toolkit (so that compilers / etc are present) but not the pieces specific to RAPIDS CI. I'd propose this would be something like mambaforge-cuda-devel, or a tag on the existing repo. It's hard to get a CUDA 11 container with both mamba/conda and CUDA compilers, and this problem has come up at least 3 times in my experience now.
(CUDA 12 is a bit simpler because the compiler and entire CUDA Toolkit can be installed via conda.)
The text was updated successfully, but these errors were encountered:
Now that CUDA 12 is fully available via conda, it might only make sense to have this for CUDA 11. I think I'm going to defer making any changes here unless/until I need it once again, but hopefully development work will shift towards CUDA 12 enough that this won't be needed at all.
As of rapidsai/ci-imgs#167, this repository is being moved into ci-imgs, so I am cleaning up this issue tracker.
I have not needed this feature for almost a year, so I am closing this issue. Also we will be dropping CUDA 11 support at some point in the future, which will negate the desire for this entirely.
I talked with @csadorf today and it seems like there might be a useful layer between
mambaforge-cuda
andci-imgs
that could include the CUDA Toolkit (so that compilers / etc are present) but not the pieces specific to RAPIDS CI. I'd propose this would be something likemambaforge-cuda-devel
, or a tag on the existing repo. It's hard to get a CUDA 11 container with both mamba/conda and CUDA compilers, and this problem has come up at least 3 times in my experience now.(CUDA 12 is a bit simpler because the compiler and entire CUDA Toolkit can be installed via conda.)
The text was updated successfully, but these errors were encountered: