Closed Galileo-Galilei closed 3 years ago
If anyone wants to try the version with 0.17 (@arnaldog12, sorry to ping directly but if you want to be a beta tester it will help ;p), feel free to install directly the plugin from the feature/compatibility-0.17 branch, it should already work and will be released within few weeks.
pip install --upgrade git+https://github.com/Galileo-Galilei/kedro-mlflow.git@feature/compatibility-0.17
Note that it does not work with kedro==0.17.1 but only with kedro==0.17.0 since the kedro core team made a breaking change between the two patch releases. I will update immediately after I release this version.
Hi, @Galileo-Galilei I'll try it in the next few days. Thanks!
Hi, @Galileo-Galilei
I have just tested the kedro-mlflow
version you sent to me in an existing project. It worked very well!
Good job! Congrats on that!
I can also take a look at the PR if you want. Just let me know. Best regards,
Thank you for the test! It would be very nice if you find time to review the code, the more eyes on it, the better the result! (note that I am still making some minor updates to the PR to improve docs and tests, but the core code should be already quite stable ).
I reopen the issue for future works, but currrently:
pip install --upgrade git+https://github.com/Galileo-Galilei/kedro-mlflow.git
Any beta tester is warmly welcome!
Description
The new
kedro==0.17.0
version is out and comes with its share of breaking changes.Context
It is important to keep the plugin up to date with kedro to avoid deprecation and increase adoption. It will also enable to adopt some highly demanded feature (e.g. #92 and maybe #124)
Possible Implementation
Most of the breaking changes comes from the
context
which is now contained in a KedroSession object. It must be updated all over the codebase.The release notes must be read in details to see if some other changes are needed.
TODO
_is_kedro_project
to check for existence of a "[tool.kedro]" section in pyproject.toml. See kedro ProjectMetaData _is_project()_get_project_globals
to retrieve info from pyproject.toml See kedro ProjectMetaData _get_project_metadata()validate_uri
which should dynamically get project_path from KedroSessionsetup
to avoid avoid passing context as argument (retrieve from KedroSession inside the function)load_context()
byKedroSession.create()
load_context()
byget_current_session()
load_context()
byKedroSession.create()