Skip to content

[data] Move non-data related stuff oustide release test timing - #65268

Open
iamjustinhsu wants to merge 2 commits into
masterfrom
jhsu/move-non-dataset-stuff-outside-timing
Open

[data] Move non-data related stuff oustide release test timing#65268
iamjustinhsu wants to merge 2 commits into
masterfrom
jhsu/move-non-dataset-stuff-outside-timing

Conversation

@iamjustinhsu

Copy link
Copy Markdown
Contributor

Description

As titled, i noticed that sometimes we have extra non-data related stuff in the benchmark_fn which can lead to variance in runtime outside of the ray data. I want to eliminate that

Related issues

Link related issues: "Fixes #1234", "Closes #1234", or "Related to #1234".

Additional information

Optional: Add implementation details, API changes, usage examples, screenshots, etc.

Signed-off-by: iamjustinhsu <jhsu@anyscale.com>

@gemini-code-assist gemini-code-assist Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

Code Review

This pull request refactors several nightly dataset benchmark scripts to ensure that only the core workload execution is timed within the benchmark runner (benchmark.run_fn). Setup steps (such as model loading, token fetching, configuration building, and datasource creation) and teardown/analysis steps (such as metrics collection, profiling delays, memory logging, and result writing) are moved outside of the timed benchmark functions. This improves the accuracy of the benchmark runtimes. I have no feedback to provide as there are no review comments.

@rayhhome rayhhome left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

LGTM! Let's run the release tests before merging to ensure correctness

Comment on lines -250 to -252
# `default_map_logical_memory_enabled` is a best practice that's required for
# Ray Data to prevent OOMs. It's not enabled by default in Ray 2.56, but we
# intend to enable it by default in a future release.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

It seems this comment was dropped accidentally; default_map_logical_memory_enabled still defaults to False for now.

Signed-off-by: iamjustinhsu <jhsu@anyscale.com>
@ray-gardener ray-gardener Bot added data Ray Data-related issues release-test release test labels Aug 7, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

data Ray Data-related issues release-test release test

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Ray fails to serialize self-reference objects

2 participants