Closed shyamcody closed 3 years ago
I would like to fix this. What you think, @shankarpandala?
Sure go-ahead
I couldn't basically figure out where the issue is generated from as both setup.py and requirements_dev.txt contains the modules I wanted to include in dependency. @brendalf seemed to know how to resolve it, so I have unassigned myself.
Working on the issue, I found this line inside the setup.py file:
python_requires=">=3.8, !=3.0.*, !=3.1.*, !=3.2.*, !=3.3.*, !=3.4.*",
This line was added on commit e3371dec from @shankarpandala. Is there any special reason for python >= 3.8? I guess that the remaining conditions can be removed, what you guys think?
It also appears that this issue happens in the latest version available, tag v0.2.7, because the setup.py there only installs the click as requirement.
setup.py on v.0.2.7:
requirements = ['Click>=7.0', ]
setup(
...
install_requires=requirements,
...
)
The setup.py inside the dev branch already fix this issue, with the commit 45fad6b1 from @shankarpandala.
setup.py on dev:
requirements = [
"click>=7.1.2",
"joblib>=0.16.0",
"lightgbm>=2.3.1",
"numpy>=1.19.1",
"optuna>=1.5.0",
"pandas>=1.0.5",
"pytest>=5.4.3",
"PyYAML>=5.3.1",
"scikit-learn>=0.23.1",
"scipy>=1.5.2",
"six>=1.15.0",
"tqdm>=4.48.0",
"xgboost>=1.1.1",
]
setup(
...
install_requires=requirements,
...
)
There still some bugs with the dependency list I'm fixing right now.
@brendalf and @shyamcody
I whole heartedly appreciate all the efforts you both are putting to make things right with lazypredict.
Please let me if there is anything I can help from my side.
Big down side is the current code is not suitable for cross validation and hyperparameter tuning. Please let me know you guys have any ideas with less efforts.
Maybe we need to incorporate some autoML tool that will take a model object in and give out the best parameters. I have already raised a separate issue for the hyperparameter optimization; and reviewing some automl frameworks as I am new in that. Will make some comments once I get a good idea. Currently, optuna and H20ai seems promising.
@brendalf Travis CI is failing for this commit stating requirements.txt not found. Can you check why it is failing even the file is there.
@shankarpandala off course. I will take a look
@shankarpandala, PR #318 should fix this issue.
Describe the bug The current dependencies mentioned are not enough. User has to manually download multiple other libraries such as tqdm,xgboost, lightgbm, pytest. So please add these to the dependency/ requirements so that it automatically downloads these files when lazypredict is installed in an environment.
To Reproduce Steps to reproduce the behavior:
Create a new python environment.
install lazypredict.
try and run lazypredict from python using the following lines:
See that module not found error will come for tqdm,xgboost, lightgbm and pytest.
Expected behavior While pip installing lazypredict, it should download all the things.
Desktop: