Skip to content
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

Look into adding support for GPU runners on CI #970

Open
cosmicBboy opened this issue Oct 18, 2022 · 4 comments
Open

Look into adding support for GPU runners on CI #970

cosmicBboy opened this issue Oct 18, 2022 · 4 comments

Comments

@cosmicBboy
Copy link
Collaborator

For #969, we need to add GPU-support for CI, since there's no way to be confident that pandera will run correctly on cudf without testing it. Relying on local testing won't be sustainable, so we need to do research on existing solutions like https://cirun.io/ .

@beckernick
Copy link

beckernick commented Oct 19, 2022

Hi! Came across this issue due to the cuDF mention. It would be wonderful for cuDF users to be able to use pandera.

Happy to try to help answer any questions that might come up!

@cosmicBboy
Copy link
Collaborator Author

@beckernick thanks! do you know how the cudf projects does unit tests on CI? I personally don't have an nvidia-enabled machine so can't test anything locally. Would love to support cudf but need CI infra in place to make sure it actually works with pandera... any tips would be much appreciated.

@beckernick
Copy link

beckernick commented Oct 19, 2022

RAPIDS projects like cuDF use a custom CI setup due to the significant GPU requirements that would likely be overkill for other projects.

At the moment, my understanding is that there are at least the following ways to do CI on GPUs with GH Actions (someone please feel free to correct me if I'm missing something):

One thing that we've seen folks do when integrating GPU support/functionality into existing projects is to run the GPU-aware CI tests on an ad-hoc basis controlled by a specific phrase used in PR comments (rather than per commit or every PR). This can be done via the issue comment event workflow trigger. If the likelihood of a given PR's interaction with GPU-relevant code is very low, this reduces resource use.

We've seen projects use cirun to streamline this with excellent results. Perhaps @aktech (cirun creator/maintainer) has some additional insight.

(cc @quasiben @shwina)

@aktech
Copy link

aktech commented Oct 19, 2022

Hey @beckernick thanks for considering cirun.io, yes you can surely use it for the above mentioned purposes and I am more than happy to help make it happen and yes it's free for open source without any restrictions. Feel free to let me know if you have any questions.

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

No branches or pull requests

3 participants