-
Notifications
You must be signed in to change notification settings - Fork 0
Add RAPIDS-Singlecell #21
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
base: main
Are you sure you want to change the base?
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
- Why not
uv
? It works fine foranndata
- If that is the case (no reason to doubt it), and
runs-on
can be configurable, this does not need to be its own.yml
file
jobs: | ||
test: | ||
name: Integration test/ test (3.x rapids-singlecell) | ||
runs-on: "cirun-aws-gpu--${{ github.run_id }}" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Is there no way to refactor the other workflow to make runs-on
a configuration setting?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I have no idea tbh.
- uses: mamba-org/setup-micromamba@v2 | ||
with: | ||
environment-file: rapids_singlecell/ci/environment.yml | ||
init-shell: >- | ||
bash | ||
post-cleanup: 'all' | ||
|
||
- name: Install rapids-singlecell | ||
id: install_pkg | ||
run: >- | ||
pip install -e .[test] | ||
"scanpy @ git+https://github.com/scverse/scanpy.git" | ||
"anndata @ git+https://github.com/scverse/anndata.git" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why do we need conda
? Why are you also installing the upstream of scanpy?
https://github.com/scverse/anndata/blob/main/.github/workflows/test-gpu.yml uses uv
, as does the rest of the tests in this repo.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I used uv for a while to do ci for rapids-singlecell. At some point every 5 runs there was an issue with installing rapids with uv so i switched back to conda
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Can we at least try it here? If not we can add an if-else
statement to use conda for rapids in the unified CI, but I've never had an issue with the anndata CI
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I know I implemented the uv ci for anndata-GPU. But you "only" install cupy and not the other rapids-stuff. I always got timeout erros from the nvidia-index
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I don’t like that this duplicates the whole workflow.
I get that if we have to use conda, that would make it necessary because the differences between workflows are too big, but are you sure that’s the only option?
How about
- you share the error message from a failed uv install with us here (I should not have to say this, you know how hard it is to debug things when users don’t give you any details)
- let us brainstorm possible solutions:
UV_HTTP_TIMEOUT
?UV_CONCURRENT_DOWNLOADS
?- manual retry?
--native-tls
?
- only if that doesn’t work, we should consider doing something totally different for one package than the others.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I get that if we have to use conda, that would make it necessary because the differences between workflows are too big, but maybe we can figure it out:
I'm not even sold on this. The difference between this and the other one really is just that micromamba line from what I can tell. Should be simple to turn it off for all the other ones, no?
id: install_pkg | ||
run: >- | ||
pip install -e .[test] | ||
"scanpy @ git+https://github.com/scverse/scanpy.git" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why install the live version of scanpy?
No description provided.