Specific Tensorflow version

[abhishektiwari]

Does benchscofi only works with tensorflow==2.4.3? Anyone using ARM architecture CPUs including Mac M1 or AWS ARM instance types can't use it.

Are you able to

1. loosen the range of package versions you've specified
2. remove package versions to allow pip attempt to solve the dependency conflict,

without impacting the features included in this package?

For JOSS Review

[RECeSS-EU-Project]

After checking out on my side, benchscofi also works for other versions of tensorflow. An alternative installation guide is

git clone https://github.com/RECeSS-EU-Project/benchscofi.git
python3 -m pip install .

where the field install_requires in setup.py is replaced by

["stanscofi",
"tensorflow",
"pulearn",
"torch",
"fastai",
"torch_geometric",
"pyFFM",
"pytorch-lightning",
"scikit-learn==1.2.*",
"libmf",]

After quite a long time, pip succeeds in solving the dependency conflict.

The reason we are not going to update setup.py is that for automated tests, such a lengthy installation time is not accepted. But this alternative approach will be added in the documentation and the README file.