fixed ci workflows#128
Conversation
|
@ryan112358 i think this is an important update related to developer experience. earlier the check system was monolith and everything seems to be sequential, now it is modular and directed acyclic graph (dag) architecture for the ci workflows here is a comparision table CI Workflow Comparison: Before vs. After
|
|
@ryan112358 i think this step should be taken into action, and if any other internal google checks are there, you can tell it here i will implement so everything will be very clear to the developer, regarding their code changes. let me know if anything need improvements, Thanks! |
@amyssnippet Anmol, this is a great refactor. Moving to a modular DAG architecture will definitely speed up the feedback loop. I ran into the dependency issue with Flax in PR #126, and having a dedicated requirements/optional deps setup for examples would have made debugging much more smoother. I support this improvement as it would add a better developer experience! 👍 |
|
@ryan112358 , i guess each and every internal checks seems to be good. so shall we proceed towards merging this soon, as this will actively improve the DX |
Great improvements, thanks for the initiative! Should be merged today, just waiting on a second internal review |

Closes #127
This PR refactors the CI pipeline to improve speed and developer experience.
Changes:
build-and-testjob into separatelint,test,matrix-tests, anddocsjobs.lintandtestrun in parallel for faster feedback.matrix-testsanddocsrun only afterlintpasses to save resources.pyproject.tomlwithoptional-dependencies(lint, test, docs) to allow granular installation.lintandtestjobs install necessary dependencies (including examples) to prevent import errors during checks.