Skip to content

Revamp cuml.accel limitations docs page#6965

Merged
rapids-bot[bot] merged 2 commits intorapidsai:branch-25.08from
jcrist:update-cuml-accel-limitations
Jul 3, 2025
Merged

Revamp cuml.accel limitations docs page#6965
rapids-bot[bot] merged 2 commits intorapidsai:branch-25.08from
jcrist:update-cuml-accel-limitations

Conversation

@jcrist
Copy link
Copy Markdown
Member

@jcrist jcrist commented Jul 3, 2025

This is a rewrite of the limitations docs page for cuml.accel. It:

  • Improves the organization and formatting to make things more readable. In particular, the headings and sidebar are now much more readable and don't line wrap in weird places.
  • Improves the uniformity across estimators. The old page had multiple authors, leading to a bit of disjoint formatting and content.
  • Ensures for every estimator we include:
    • When we fallback to CPU and why (and that this matches the implementation).
    • What fitted attributes are currently missing
    • What methods will always fallback to CPU
    • How to compare results between sklearn and cuml.accel (when relevant).
  • Also added some general guidance on when and why a model might fallback.

Fixes #6956.

This is a rewrite of the limitations docs page for `cuml.accel`. It:

- Improves the organization and formatting to make things more readable.
  In particular, the headings and sidebar are now much more readable and
  don't line wrap in weird places.
- Improves the uniformity across estimators. The old page had multiple
  authors, leading to a bit of disjoint formatting and content.
- Ensures for every estimator we include:
  - When we fallback to CPU and why (and that this matches the
    implementation).
  - What fitted attributes are currently missing
  - What methods will always fallback to CPU
  - How to compare results between sklearn and cuml.accel (when
    relevant).
- Also added some general guidance on when and why a model might
  fallback.
@jcrist jcrist self-assigned this Jul 3, 2025
@jcrist jcrist added doc Documentation non-breaking Non-breaking change labels Jul 3, 2025
@jcrist
Copy link
Copy Markdown
Member Author

jcrist commented Jul 3, 2025

Going back over every wrapped estimator was a useful exercise - I've found a few cases where our logic was incorrect (and we should be falling back). This docs page matches what the code currently does - once this is in, I'll push a small followup PR fixing a few improperly handled hyperparams (and updating the doc accordingly).

@jcrist jcrist requested a review from csadorf July 3, 2025 20:20
Copy link
Copy Markdown
Contributor

@csadorf csadorf left a comment

Choose a reason for hiding this comment

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

Great improvement! Just a few minor change requests.

Comment thread docs/source/zero-code-change-limitations.rst Outdated
Comment thread docs/source/zero-code-change-limitations.rst Outdated
Comment thread docs/source/zero-code-change-limitations.rst Outdated
Comment thread docs/source/zero-code-change-limitations.rst Outdated
Comment thread docs/source/zero-code-change-limitations.rst Outdated
Comment thread docs/source/zero-code-change-limitations.rst
@jcrist
Copy link
Copy Markdown
Member Author

jcrist commented Jul 3, 2025

Thanks, I believe I've addressed all points.

Copy link
Copy Markdown
Contributor

@csadorf csadorf left a comment

Choose a reason for hiding this comment

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

🚢

@csadorf
Copy link
Copy Markdown
Contributor

csadorf commented Jul 3, 2025

/merge

@rapids-bot rapids-bot Bot merged commit cc30652 into rapidsai:branch-25.08 Jul 3, 2025
54 checks passed
@jcrist jcrist deleted the update-cuml-accel-limitations branch July 3, 2025 21:36
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

doc Documentation non-breaking Non-breaking change

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Improve documentation of cuml.accel limitations

3 participants