Make the docs build warning-free under --strict and gate it in CI (Issue #241)#265
Open
egmaminta wants to merge 3 commits into
Open
Make the docs build warning-free under --strict and gate it in CI (Issue #241)#265egmaminta wants to merge 3 commits into
egmaminta wants to merge 3 commits into
Conversation
decord ships wheels only for manylinux2010_x86_64 and win_amd64 (no macOS arm64 wheel and no sdist), so the previous unconditional dependency broke uv sync on macOS. decord is now gated to non-darwin platforms and av (PyAV) is pulled in on darwin instead. VideoDataset selects a decoder at runtime: decord where available, otherwise a small PyAV-backed reader that exposes the same integer-indexing and get_batch API and returns identical uint8 (N, C, H, W) tensors. MP4 writing already went through imageio. Adds an av round-trip test (skipped when av is absent) and updates the backend import-error test.
mkdocs build --strict was failing with 13 warnings, but CI only ran a plain mkdocs build, so these could (and did) slip through unnoticed. - lance: the LanceDataset Args block lumped six parameters onto a single line, which griffe cannot parse into individual parameter entries. Document each one separately so the Dataset reference renders correctly. - data/utils, data/normalization: add the parameter and return type annotations griffe flagged (load_dataset, convert, column_normalizer, get_scaler). Forward references stay under TYPE_CHECKING. - dmc: the expert-training link pointed at scripts/expert/train_policies.py, which does not exist. Point it at the SAC trainer actually used, scripts/expert/sac_online.py. - nav: add envs/piecewise.md, a finished page that was not linked anywhere. With those resolved, switch the docs CI job to mkdocs build --strict so the reference stays warning-free.
lucas-maes
reviewed
Jun 22, 2026
Comment on lines
+16
to
+20
| if TYPE_CHECKING: | ||
| from stable_pretraining.data.transforms import WrapTorchTransform | ||
|
|
||
| from stable_worldmodel.data.dataset import Dataset | ||
|
|
Collaborator
|
thank you for the pr! can you address my comment? |
Contributor
Author
Good day to you @lucas-maes ! So these two imports back the type annotations added to the autodoc'd functions in this PR:
They have to be resolvable because And since this PR switches CI to
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
While going through the docs for Issue #241 I noticed
mkdocs build --strictfails with 13 warnings, but CI only runs a plainmkdocs build, so they ship silently. This PR clears all of them and turns on--strictso the API reference stays clean.LanceDatasetdocumented six params on one line, so griffe couldn't split them into individual entries and they rendered wrong. Each is documented separately now.data/utils,data/normalizationadded the parameter/return annotations griffe flagged (load_dataset,convert,column_normalizer,get_scaler); forward refs kept underTYPE_CHECKING.scripts/expert/train_policies.py, which doesn't exist; now points at the SAC trainer actually used,scripts/expert/sac_online.py.envs/piecewise.md, a finished page that wasn't reachable.Verified these locally by running
mkdocs build --strict(clean) pluspytest tests/data(passes), ruff also clean.Refs #241: I'm open to tackle data-layer docstrings, a beginner concepts page, or the solver reference . Just let me know.