Port to Pandas 2.0

[hasan7n]

Describe the bug
when running gandlf_run, I am encountering:

ERROR: 'DataFrame' object has no attribute 'append'

To Reproduce

Train a model using gandlf_run.

I trained using 2d_rad_segmentation data from https://upenn.box.com/shared/static/y8162xkq1zz5555ye3pwadry2m2e39bs.zip and the config file from samples in the repo config_classification.yaml

Additional context

• check the changelog of pandas here
• The training runs successfully when downgrading pandas to 1.5.3

[sarthakpati]

Unless we have a specific requirement for pandas>=2.0, my recommendation would be to stick with pandas<2.0. This will ensure long-term stability of the next release. We can update this requirement along with PyTorch (#587) in the future.

[sarthakpati]

Reopening and assigning this for code-reorg.

[github-actions]

Stale issue message

[github-actions]

Stale issue message

[Ainesh06102004]

I'll gladly take this on.

[sarthakpati]

Thanks @Ainesh06102004, this is now assigned to you!

[Ainesh06102004]

When upgrading pandas to 2.0.0 version, it is becoming incompatibe with numpy 1.22.0 version, which is currently being used.

[sarthakpati]

When upgrading pandas to 2.0.0 version, it is becoming incompatibe with numpy 1.22.0 version, which is currently being used.

@szmazurek you are working on the numpy version port, right?

[szmazurek]

Yes, if nothing goes wrong the PR goes tomorrow