Reland "Add experimental infer_stypes(_low_cardinality: bool)"

[akihironitta]

Relands #987, which #988 reverted together with the AlignCategories removal.

• Restore the experimental infer_stypes(_low_cardinality=...) from Add experimental infer_stypes(_low_cardinality: bool) #987 unchanged.
• Enable it for SDMKumoTabularModel in the TabArena and BeyondArena adapter through a low_cardinality class attribute, as AutoGluon's feature generator hands binary columns over as integers. Other adapters keep "off".
• Drop the category pinning in _preprocess from Add experimental infer_stypes(_low_cardinality: bool) #987: with AlignCategories(sort_by="value") back in the KumoTabular recipe (Reverts AlignCategories removal from TabArena and BeyondArena benchmarks #988), categories are fitted on the training data in value order and unseen values become missing.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[coderabbitai]

Navigate logical layers of code changes, visualize relationships, and explore their blast radius.

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository: NVIDIA/structured-data-models/.coderabbit.yaml

Review profile: QUIET

Plan: Enterprise

Run ID: 6756968d-53e6-4da7-9d20-a1d847fbee70

📥 Commits

Reviewing files that changed from the base of the PR and between 0cab2be and 725dbf8.

📒 Files selected for processing (3)

• benchmark/tabular/model.py
• sdm/stype.py
• test/test_stype.py

Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 11 remain after this review.

---

📝 Summary

Summary by CodeRabbit

• New Features
  • Kumo Tabular models now identify numeric columns with two or three distinct values as categorical when the table contains at least 151 rows. This applies to integer and floating-point columns, as well as supported decimal columns.
• Bug Fixes
  • Standard SDM inference continues to classify numeric columns as numerical by default, preserving existing behavior.

Walkthrough

infer_stypes can optionally classify numeric columns with two or three distinct values as categorical when they have at least 151 rows. SDMModel defaults this option off, while SDMKumoTabularModel enables it. Tests cover pandas, Arrow, and cuDF.

Changes

Low-cardinality numeric inference

Layer / File(s)	Summary
Inference policy and backend classification sdm/stype.py, test/test_stype.py	infer_stypes passes the policy to the Arrow, pandas, and cuDF helpers. When enabled, numeric columns qualify as categorical if they have at least 151 rows and two or three distinct values. Tests cover the row threshold and cardinalities.
Model inference settings benchmark/tabular/model.py	SDMModel defaults the setting to off and passes it during fitting. SDMKumoTabularModel sets it to infer.

Priority: ⬇️ Low

Estimated code review effort: 3 (Moderate) | ~20 minutes

Merge Risk: ⚪ Minimal · up to 725db

No confirmed issue blocks merging. Run the normal backend tests before merging.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 8.33% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 12 functions across 3 files.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly identifies the main change: relanding experimental low-cardinality inference in infer_stypes.
Description check	✅ Passed	The description accurately explains the reland, adapter-specific enablement, and category preprocessing behavior.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

• Fix all pre-merge checks with AI

✨ Finishing Touches 💡 1

📝 Generate docstrings 💡

• Commit to this branch
• Create a new PR

🧪 Generate unit tests (beta)

• Commit to this branch
• Create a new PR

---

_{Comment @coderabbitai help to get the list of available commands.}