Uv and Ray: Pain-Free Python Dependencies in Clusters
8 comments
·June 27, 2025lz400
rsfern
Are there particular libraries that make your setup difficult? I just manually set the index and source following the docs (didn’t know about the auto index feature) and pin a specific version if I really have to with `uv add “torch==2.4”`. This works pretty well for me for projects that use dgl, which heavily uses C++ extensions and can be pretty finicky about working with particular versions
This is in a conventional HPC environment, and I’ve found it way better than conda since the dependency solves are so much faster and I no longer experience PyTorch silently getting downgraded to cpu version of I install a new library. Maybe I’ve been using conda poorly though?
Kydlaw
You should give a try to https://pixi.sh/latest/ (I am not involve in the project).
They are a little more focus on scientific computing than uv, which is more general. They might be a better option in your case.
miohtama
Would it be possible to use Docker to manage native dependencies?
devjab
lz400
the problem is that you still need to install all the low level stuff manually, conda does it automatically
gcarvalho
I was pleasantly surprised to try the guide out and see that it just worked:
λ uv venv && uv pip install torch --torch-backend=auto
λ uv run python -c 'import torch; print(torch.cuda.is_available())'
True
jessekv
It would be a fun callback if the demo was a factorial server:
https://joearms.github.io/#2013-11-21%20My%20favorite%20Erla...
Unfortunately uv is usually insufficient for certain ML deployments in Python. It's a real pain to install pytorch/CUDA with all the necessary drivers and C++ dependencies so people tend to fall back to conda.
Any modern tips / life hacks for this situation?