Pre-docs cleanup: fix the typo'd names and tighten the lint gate

[kozlov721]

Purpose

Prepare the repository for the Google docstyle migration and clean it
up in general. The PR mirrors the pre-docs cleanup in modelconverter
(luxonis/modelconverter#276) and the "Small breaking changes" PR in
luxonis-ml (luxonis/luxonis-ml#426). The docstyle migration follows
in a separate PR; it converts the docstrings that this PR makes
correct.

Specification

• Add the typos pre-commit check with the same pin as the sibling
  repositories, and fix the 75+ typos it found.
• Rename the typo'd public names:
  • reparametrize → reparameterize (base class, module file,
    GeneralReparameterizableBlock, methods).
  • AttentionRefinmentBlock → AttentionRefinementBlock.
  • mask_coeficients → mask_coefficients (head output key, loss
    parameter, tests — renamed together, so the wiring holds).
  • PPLCNetV3 detecion_blocks → detection_blocks.
• Expand the ruff configuration to the modelconverter rule set:
  ASYNC, EXE, G, PGH, SLOT, full PERF, formatted docstring
  code examples, and a ban on __future__ imports. Fix the new
  findings and bump the ruff hook to match the ~=0.16 dev pin.
• Fix all 49 pydoctor "documented parameter does not exist" warnings.
• Correct the stale paths and examples in the README.
• Exclude the stale build directory from pyright.
• Anchor the head-branch regexes in .github/labeler.yaml, so a
  branch such as ci/release-automation no longer gets the release
  label and no longer disappears from the release notes.
• Bump the prettier hook from v3.8.3 to v3.9.6.
• Format the Python code blocks in README.md and examples/README.md.
• Fix the "wit ha single key" typo in a node warning. The typos hook
  does not find it, because both words are correct.
• Stop the variant metaclass from calling __init__ inside its
  except NotImplementedError handler. A node without variants often
  builds itself from a remote checkpoint. A network failure there
  chained onto the NotImplementedError, so the traceback opened with
  a variants error and hid the real cause.
• Add unit tests for the bounding-box helpers, the tuner parameter
  parsing, and the metadata decode pass-through.

Out of scope, with their own follow-ups:

• The remaining pydoctor warnings are format-level (@license fields,
  link targets). The docstyle migration rewrites those docstrings.
• The typing debt (Any, cast, blanket type-ignores; PGH003 sits
  in the ignore list for now), like chore/type-gate-and-coverage in
  modelconverter.

Dependencies & Potential Impact

• Breaking: the detecion_blocks attribute of PPLCNetV3 is a
  registered submodule, so its state-dict keys change. Old PPLCNetV3
  checkpoints do not load any more. The node has no remote weights
  URL, so only local checkpoints are affected.
• Breaking: code that imports the renamed classes or calls the
  renamed methods must update the names. Configs are not affected; no
  registry name changes.
• No dependency changes.

Deployment Plan

None / not applicable — the changes ship with the next release.

Testing & Validation

• prek run --all-files: all hooks pass, including the new typos
  hook.
• pydoctor --docformat=epytext: the 49 content warnings go to 0.
• pyright --warnings: at exact parity with main (2 pre-existing
  aimet_torch import errors from the optional extra).
• pytest -m unit: 547 passed. The 9 failures are the known expired
  gcloud-ADC failures in tests/unittests/test_losses, not code
  failures.

AI Usage

Assisted-by: Claude:claude-fable-5

Submitted code was reviewed by a human: NO

The author is taking the responsibility for the contribution: YES

🤖 Generated with Claude Code

https://claude.ai/code/session_01KXj7GZFCcdShtaTweRNH9j

Summary by CodeRabbit

• Documentation

  • Corrected typos, parameter names, type descriptions, links, examples, and error messages across documentation and command references.
  • Updated model and configuration terminology for clearer guidance.

• Refactor

  • Standardized reparameterization terminology across model components and export workflows.
  • Corrected public class, method, configuration, and result-key names for consistency.

• Tests

  • Expanded coverage for tuning parameters, detection utilities, metadata decoding, and segmentation losses.

• Chores

  • Improved linting and spelling checks, refreshed formatting rules, and refined automated branch labeling.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Advanced

Run ID: 6e9e8d22-65dc-4bac-bd02-58a4120a26f3

📥 Commits

Reviewing files that changed from the base of the PR and between 95e37de and 49bb738.

📒 Files selected for processing (5)

• luxonis_train/core/utils/tune_utils.py
• luxonis_train/utils/boundingbox.py
• luxonis_train/variants.py
• tests/unittests/test_utils/test_boxutils.py
• tests/unittests/test_variants.py

Included review availability: Your plan provides up to 1 included review per hour; 0 remain after this review.

---

📝 Walkthrough

Walkthrough

The pull request standardizes spelling, naming, documentation, lint tooling, branch-label patterns, and test coverage. It also centralizes parameter-count formatting and renames the reparameterization API.

Changes

Repository consistency and naming cleanup

Layer / File(s)	Summary
Tooling and documentation updates .github/labeler.yaml, .pre-commit-config.yaml, pyproject.toml, README.md, examples/README.md	Branch patterns, pre-commit hooks, Ruff and Pyright settings, Typos settings, links, and examples are updated.
Public names and model references luxonis_train/nodes/blocks/*, luxonis_train/nodes/base_node.py, luxonis_train/lightning/luxonis_lightning.py, luxonis_train/nodes/backbones/*, luxonis_train/nodes/necks/*	Reparameterization and attention block names are corrected. Imports, exports, calls, backbone references, and segmentation result keys use the corrected names.
Model and utility implementation updates luxonis_train/callbacks/luxonis_model_summary.py, luxonis_train/core/utils/tune_utils.py, luxonis_train/utils/boundingbox.py, luxonis_train/variants.py	Parameter-count formatting is centralized. Optuna sampling uses suggest_float. Non-maximum suppression uses index-based filtering. Variant error handling is restructured.
Documentation and validation luxonis_train/**, tests/unittests/*	Docstrings, comments, annotations, local names, error messages, and tests are aligned with existing interfaces and behavior.

Estimated code review effort: 3 (Moderate) | ~25 minutes

Merge Risk: 🔵 Low · up to 49bb7

The update modernizes tuning APIs and adds coverage, but the new tests still exercise dependency APIs with a documented removal path. This is a bounded compatibility concern requiring owner awareness before merge.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 53.28% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 122 functions across 56 files.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly summarizes the main cleanup work, including typo and public-name corrections and stricter lint checks.
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 💡

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch chore/pre-docs-cleanup

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

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

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.79%. Comparing base (795f485) to head (49bb738).
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #443      +/-   ##
==========================================
+ Coverage   94.61%   94.79%   +0.18%     
==========================================
  Files         290      291       +1     
  Lines       15868    15941      +73     
==========================================
+ Hits        15013    15112      +99     
+ Misses        855      829      -26     

Flag	Coverage Δ
pytest	94.79% <100.00%> (+0.18%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.