The problem described inside the issues was already fixed by the commit 45fad6b from @shankarpandala.
But, there are still some problems related.
1. Dependency list inside the setup.py file
This approach breaks the principle of single responsibility, making the setup.py more extensive that it should be.
The dependency list are now listed inside the requirements.txt file, that is the python community default.
2. Dependency packages versions
All the packages were configured to install the most recent version available, versions that are not tested with lazypredict and could cause bugs in the future.
For example:
The scikit-learn were installing the version 0.24 and that makes lazypredict crash, because some models and functions of sklearn that lazypredict uses are now deprecated (are still there, but are imported from a different module).
To fix this and to prevent new bugs in the future, the packages versions are now fixed.
3. Extra
I fixed a typo inside LazyRegressor docstring (clf -> reg).
Lazypredict requires python ">=3.8, !=3.0.*, !=3.1.*, !=3.2.*, !=3.3.*, !=3.4.*", described inside the setup.py.
I don't know why it should requires python >= 3.8 instead of 3.6 or 3.7, but the remaining conditions "!=3.0.*, !=3.1.*, !=3.2.*, !=3.3.*, !=3.4.*" are not necessary, so I removed.
Closes #284.
The problem described inside the issues was already fixed by the commit 45fad6b from @shankarpandala. But, there are still some problems related.
1. Dependency list inside the setup.py file This approach breaks the principle of single responsibility, making the setup.py more extensive that it should be. The dependency list are now listed inside the requirements.txt file, that is the python community default.
2. Dependency packages versions All the packages were configured to install the most recent version available, versions that are not tested with lazypredict and could cause bugs in the future.
For example: The scikit-learn were installing the version 0.24 and that makes lazypredict crash, because some models and functions of sklearn that lazypredict uses are now deprecated (are still there, but are imported from a different module).
To fix this and to prevent new bugs in the future, the packages versions are now fixed.
3. Extra
clf
->reg
).">=3.8, !=3.0.*, !=3.1.*, !=3.2.*, !=3.3.*, !=3.4.*"
, described inside the setup.py. I don't know why it should requires python >= 3.8 instead of 3.6 or 3.7, but the remaining conditions"!=3.0.*, !=3.1.*, !=3.2.*, !=3.3.*, !=3.4.*"
are not necessary, so I removed.