🔒 🤖 CI Update lock files for array-api CI build(s) 🔒 🤖 #29576
Conversation
41b06a7 to
8f91a46
Compare
|
@lesteve @betatim I had to manually relabel this PR as the original label set by the @github-actions bot did not trigger the CUDA GPU CI workflow on the original 8f91a46 commit for some reason. After pushing an empty commit, syncing with https://github.com/scikit-learn/scikit-learn/actions/runs/10141110347/job/28037662041?pr=29576 Any idea why the first label by @github-actions was not enough? |
Yep I have seen the same as well last week, no idea why, I remove and added the CUDA CI label manually which worked #29537 (comment) ... |
|
Good question. Removing and adding the label is what will trigger the workflow. It is configured to react to the "label" event, so new commits don't help. I have no good idea right away why it didn't get triggered. I tried to find a log of all the events related to the repository and/or a way each time a workflow is triggered and evaluates the |
…29576) Co-authored-by: Lock file bot <noreply@github.com> Co-authored-by: Olivier Grisel <olivier.grisel@ensta.org>
|
Bumped into this today, maybe the fact that adding the label "CUDA CI" does not trigger the CUDA CI run is because of triggering a workflow from a workflow limitations, see doc. Apparently you need a PAT rather than the default GITHUB_TOKEN when you are in this case. The last example in the doc is very similar to our use case I think:
I don't remember if we have a PAT for some of the CI that we could try to see whether that fixes the issue ... |
|
I created #29781 to track this, because it took me a while to find the relevant PRs where we discussed this 😅 |
Update lock files.
Note
If the CI tasks fail, create a new branch based on this PR and add the required fixes to that branch.