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] Fix ResourceChangingScheduler dropping PGF args #30304

Merged

Conversation

Yard1
Copy link
Member

@Yard1 Yard1 commented Nov 15, 2022

Signed-off-by: Antoni Baum [email protected]

Why are these changes needed?

ResourceChangingScheduler didn't carry over attributes from the base PlacementGroupFactory, causing the resources to be constantly reallocated as they failed the equality check between the old resources and new resources.

Related issue number

Closes #30265

Checks

  • I've signed off every commit(by using the -s flag, i.e., git commit -s) in this PR.
  • 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 :(

Copy link
Contributor

@amogkam amogkam left a comment

Choose a reason for hiding this comment

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

ResourceChangingScheduler changes lgtm.

Can we do the GBDTTrainer changes in a separate PR?

I'd like to see if we can simplify the logic and keep a linear conversion path, rather than the current branched conversion path.

Instead of this:


Scaling Config -> PGF--------------------->Combine -> ...
                                          ^
Scaling Config -> Ray Params -> PGF-------|

Ideally we can just have this

Scaling Config -> PGF

Or at least this

Scaling Config -> Ray Params -> PGF

@Yard1
Copy link
Member Author

Yard1 commented Nov 18, 2022

Converting a ScalingConfig to PGF directly was the first idea, but the issue is that RayParams for LightGBM has some extra logic/defaults and we don't have an API right now to do the same for ScalingConfig.

Refactored and split into #30470

@amogkam amogkam merged commit 9f199bf into ray-project:master Nov 18, 2022
@Yard1 Yard1 deleted the fix_resource_changing_scheduler_pgf_args branch November 18, 2022 19:12
WeichenXu123 pushed a commit to WeichenXu123/ray that referenced this pull request Dec 19, 2022
…#30304)

ResourceChangingScheduler didn't carry over attributes from the base PlacementGroupFactory, causing the resources to be constantly reallocated as they failed the equality check between the old resources and new resources.

Signed-off-by: Antoni Baum <[email protected]>
Signed-off-by: Weichen Xu <[email protected]>
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.

[AIR] ResourceChangingScheduler causes tuning to hang with a large trial number
3 participants