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

[tune] Enforce one future at a time for any given trial at any given time. #20783

Merged
merged 5 commits into from
Dec 3, 2021

Conversation

xwjiang2010
Copy link
Contributor

@xwjiang2010 xwjiang2010 commented Nov 30, 2021

Why are these changes needed?

Also enforce disabling (instead of allowing user to override this) buffer training when checkpoint_at_end is used.

Related issue number

Checks

  • I've run scripts/format.sh to lint the changes in this PR.
  • I've included any doc changes needed for https://docs.ray.io/en/master/.
  • I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
  • Testing Strategy
    • Unit tests
    • Release tests
    • This PR is not tested :(

@xwjiang2010 xwjiang2010 changed the title [try out] One future at a time for any given trial at any given time. [tune] Enforce one future at a time for any given trial at any given time. Dec 1, 2021
@xwjiang2010 xwjiang2010 added this to the Tune Tech Debt Reduction milestone Dec 1, 2021
Copy link
Contributor

@krfricke krfricke left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks good but please run the result throughput single node and cluster release tests to make sure they still pass (you may have to set the environment variable there)

Comment on lines +590 to +592
assert len(
out
) <= 1, "Expecting one future for any given trial at any given time."
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Possibly unrelated to this PR: Should we rename this into find_training_future as we only use this helper to find futures in self._running?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, and instead of retuning a list, we should just return the item.
May do that separately tho. Want to first enforce this basic behavior first.

@xwjiang2010
Copy link
Contributor Author

Added env variable for result throughput release tests.

@xwjiang2010
Copy link
Contributor Author

@krfricke krfricke merged commit 368da17 into ray-project:master Dec 3, 2021
@xwjiang2010 xwjiang2010 deleted the find_item branch July 26, 2023 19:50
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

Successfully merging this pull request may close these issues.

2 participants