Reverts AlignCategories removal from TabArena and BeyondArena benchmarks

[akihironitta]

Reverts these:

• Add experimental infer_stypes(_low_cardinality: bool) #987
• Avoid running unnecessary AlignCategories #973

Closes #986.

[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: 988cec32-bcc7-4ad7-b4f5-2dcedfe56847

📥 Commits

Reviewing files that changed from the base of the PR and between 1dbc7af and 6a9174d.

📒 Files selected for processing (3)

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

💤 Files with no reviewable changes (1)

• 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

• Behavior Changes
  • Integer, floating-point, and decimal columns are now inferred as numerical regardless of how many distinct values they contain. Low-cardinality detection is no longer available as an inference option.
  • Tabular model adapters no longer pin low-cardinality integer categories to values seen during training. Category alignment remains in categorical feature and target processing, which may affect how categories are handled when fitting models.

Walkthrough

The changes remove low-cardinality classification from numeric type inference and remove Kumo-specific category handling from tabular benchmark preprocessing. Fitting now obtains the default recipe directly.

Changes

Type inference and benchmark preprocessing

Layer / File(s)	Summary
Numeric type inference sdm/stype.py, test/test_stype.py	The inference API and backend helpers no longer use a low-cardinality policy. Numeric columns are inferred as numerical. Low-cardinality detection helpers and their test are removed.
Tabular benchmark preprocessing benchmark/tabular/model.py	Fitting calls default_recipe() directly. Training preprocessing no longer receives is_train=True. The Kumo-specific preprocessing and recipe overrides are removed.

Priority: ⬇️ Low

Estimated code review effort: 2 (Simple) | ~15 minutes

Merge Risk: ⚪ Minimal · up to 6a917

No actionable merge-blocking issue was established. The effect of omitting the training flag remains unverified.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 11.11% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 9 functions across 2 files.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
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.
Title check	✅ Passed	The title clearly states that the pull request restores AlignCategories in the TabArena and BeyondArena benchmarks. This matches the main change.
Description check	✅ Passed	The description identifies the reverted pull requests and the related issue. It is directly related to the changeset.

• 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.}