contributing, development setup, error message in tests

[asgalon]

To learn the steps necessary to contribute fixes and improvements, I have now started to work on ticket #1221.
This is in module 02

What I have done until now, following tinytorch/CONTRIBUTING.md:

• created local feature branch feature/feature-1221

python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate
pip install -r requirements.txt
pip install -e .
tito --version       # Check TinyTorch version
tito system health   # Verify environment
tito module status   # See module progress

• finished module 01

• edited src/02_activation/02_activation_py to suppress the warning (I only switched the warning in the test off and added a note to the fact. so no change of the actual activation module results)

• gone through the modified module in the notebook to verify the correct output, then completed the module

then:

pytest tests/
tito module test -v 02

Now in the last step, I see the following error message:

────────────────────────────────────────────────────────────────────────────────── Phase 2: Module Tests (with educational output) ───────────────────────────────────────────────────────────────────────────────────
Running pytest with WHAT/WHY context for each test...

ERROR: usage: __main__.py    [...]
__main__.py: error: unrecognized arguments: --tinytorch
  inifile: /Users/peter/PycharmProjects/cs249r_book/tinytorch/pyproject.toml
  rootdir: /Users/peter/PycharmProjects/cs249r_book/tinytorch


✗ Phase 2 FAILED: Module tests

The parameter is set in tinytorch/tito/commands/module/test.py, not sure what is going wrong there.

If I run pytest tests/02_activations everything is ok, apart from numerous warnings from integration tests that do not seem to have anything to do with the change I made_

================================================================================================== warnings summary ==================================================================================================
tests/integration/test_api_simplification_integration.py::test_pytorch_api_compatibility
  /Users/peter/PycharmProjects/cs249r_book/tinytorch/.venv/lib/python3.12/site-packages/_pytest/python.py:170: PytestReturnNotNoneWarning: Test functions should return None, but tests/integration/test_api_simplification_integration.py::test_pytorch_api_compatibility returned <class 'dict'>.
  Did you mean to use `assert` instead of `return`?
  See https://docs.pytest.org/en/stable/how-to/assert.html#return-not-none for more information.
    warnings.warn(

[31 more similar warnings...]

So I am not quite sure what to do for a satisfactory pre-commit test result here, if it is some problem with the local project setup or something missing...

Also, the following is probably not completely up to date, shouldn't I prepare a PR instead? Up until now, I have been working completely with a fork as remote push target to avoid spilling over any onboarding mistakes....

# 6. Merge to dev when complete
git checkout dev
git merge feature/your-improvement
git branch -d feature/your-improvement

Update: The error is independent from the module tested, so it most probably is some tool issue. it is a bit difficult to make out exactly what part is failing there. Maybe some version update issue?

[profvjreddi]

Thanks @asgalon, and thanks for working on #1221! Your PRs have been really helpful (both #1262 and #1271 are merged).

The --tinytorch error you hit was a real bug. The flag was being passed by tito module test but never registered in conftest.py (a plugin that handled it got removed during cleanup and nobody wired it back up). This is now fixed in commit 93de573. After pulling the latest, tito module test should work.

The integration test warnings about PytestReturnNotNoneWarning are harmless, those tests are returning dicts for reporting purposes and pytest is just being noisy about it. Nothing to worry about for your contribution.

And yes, a PR from your fork is the right workflow. The CONTRIBUTING.md instructions about merging locally are outdated. Fork, branch, PR is the way to go.

Really appreciate you taking the initiative to work through the contributor flow and documenting the rough edges.