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

[Discussion] The future of Trial metadata parameters. #2390

Open
Electronic-Waste opened this issue Jul 17, 2024 · 5 comments
Open

[Discussion] The future of Trial metadata parameters. #2390

Electronic-Waste opened this issue Jul 17, 2024 · 5 comments
Labels
help wanted Extra attention is needed kind/discussion

Comments

@Electronic-Waste
Copy link
Member

Electronic-Waste commented Jul 17, 2024

As we discuss in issue #2369 , we choose to always pass the Trial name as env variable KATIB_TRIAL_NAME to the training container. And we implemented this feature as function mutatePodEnv and used it here:

// Add Katib Trial labels to the Pod metadata.
mutatePodMetadata(mutatedPod, trial)

Here are some discussions about it: #2369 (comment) #2369 (comment) #2369 (comment)

We use this issue to track the discussion. Any suggestions are welcome!

Copy link

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

@andreyvelich
Copy link
Member

/remove-lifecycle stale
/good-first-issue

@mickvangelderen
Copy link

mickvangelderen commented Jan 10, 2025

In my project I would like to know the experiment name and trial name:

KATIB_EXPERIMENT_NAME
KATIB_TRIAL_NAME

as these values may be generated by K8S.

It would be nice to also know the namespace:

KATIB_NAMESPACE

as well, even though that value may be known.

For my project, I also need

KATIB_BASE_URL

but I suppose Katib does not necessarily know it's external-facing base URL and so I can inject that value into the container envs myself.

I compute the experiment url and trial url from the base URL, namespace and names. If the URLs are not stable, we can consider also exposing the relative URLs of the experiment and trial:

KATIB_RELATIVE_EXPERIMENT_URL
KATIB_RELATIVE_TRIAL_URL

Not sure that's worth doing or how they should be named. Currently I am not sure how to construct the trial url though, see #2474.

@Electronic-Waste
Copy link
Member Author

Could you please pin this issue so that more people can join the discussion? @andreyvelich
/remove-good-first-issue
/kind discussion

@google-oss-prow google-oss-prow bot added kind/discussion and removed good first issue Good for newcomers labels Jan 11, 2025
@Electronic-Waste
Copy link
Member Author

I think it's a good idea to pass corresponding experiment name and the namespace to the env of Pod! As for the url, I'm not sure if we need to pass it to the pod.

WDYT👀 @kubeflow/wg-automl-leads @tariq-hasan @helenxie-bit

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
help wanted Extra attention is needed kind/discussion
Projects
None yet
Development

No branches or pull requests

3 participants