Skip to content

Conversation

jrbourbeau
Copy link
Contributor

This should pull in the right GPU version of pytorch from conda-forge by default now, so we can remove older workarounds

@jrbourbeau
Copy link
Contributor Author

Note I've not tested this yet, so will want to do so before closing #53

@ntabris
Copy link
Member

ntabris commented May 12, 2025

@jrbourbeau can you say more / remind us why you'd expect this to pull in GPU version of pytorch?

@jrbourbeau
Copy link
Contributor Author

This is not working as expected (currently it's pulling in the CPU version)

A bare conda install -c conda-forge pytorch will now install the GPU compiled version if run on a machine with a GPU. I was hoping that would work inside the RAPIDS image but isn't

@jrbourbeau jrbourbeau changed the title Fix GPU docker image [WIP] Fix GPU docker image May 13, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants