Pass `CONDA_OVERRIDE_CUDA` to `with_cuda` of conda-lock

nkaretnikov commented 9 months ago

Fixes #719.

Description

This pull request makes it possible to set the CUDA version by passing the value of the CONDA_OVERRIDE_CUDA specification variable to the with_cuda parameter of conda-lock.

See https://docs.conda.io/projects/conda/en/latest/user-guide/tasks/manage-virtual.html#overriding-detected-packages

How I tested this:

Created this test env via the server admin environment page because conda-store UI strips variables even in raw YAML mode, so it doesn't work there:

channels:
- conda-forge
dependencies:
- pytorch
- ipykernel
- pip:
  - nothing
description: test2
name: test3
prefix: null
variables:
  CONDA_OVERRIDE_CUDA: '12.0'

Checked that the pytorch package in the generated lockfile has constraints matching this version of cuda. The url of the pytorch package should also reflect that, e.g., https://conda.anaconda.org/conda-forge/linux-64/pytorch-2.1.0-cuda120py312hfe5e8c6_301.conda.
Additionally, on a machine with an NVIDIA card that's configured to use CUDA:

% CONDA_STORE_USERNAME=test CONDA_STORE_PASSWORD=password  python -m conda_store --conda-store-url="http://localhost:8080/conda-store/" --auth=basic run test/test3:4 -- python
>>> import torch
>>> torch.__file__
'/tmp/conda-store/4/lib/python3.12/site-packages/torch/__init__.py'
>>> print(torch.version.cuda)
12.0

This version matches the one in the lockfile, which comes from the variable. Note: I had to set the url when calling conda-store CLI in order to match the url used by the server when running via docker.

Repeated the same with versions 10.0 and 11.0. Observed that these affect constraints in the pytorch package in the lockfile. Note that I've used different major versions as minor versions don't affect the constraints in the lockfile.

Pull request checklist

[x] Did you test this change locally?
[ ] Did you update the documentation (if required)?
[ ] Did you add/update relevant tests for this change (if required)?

Additional information

netlify[bot] commented 9 months ago

Deploy Preview for kaleidoscopic-dango-0cf31d ready!

Name	Link
Latest commit	8c9390c2c538cafd68b6080f9368e1b5af12d5e1
Latest deploy log	https://app.netlify.com/sites/kaleidoscopic-dango-0cf31d/deploys/65a6ec80fa29ce00088fc76f
Deploy Preview	https://deploy-preview-721--kaleidoscopic-dango-0cf31d.netlify.app
Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

To edit notification comments on pull requests, go to your Netlify site configuration.

nkaretnikov commented 9 months ago

@amjames Do you have cycles to review this one?

nkaretnikov commented 9 months ago

@trallard Could you approve so I could merge? Andrew has reviewed this.

trallard commented 8 months ago

Added a non-blocking comment otherwise we can merge

conda-incubator / conda-store