Add weights and biases support

[nictru]

This adds a first draft of what native weights and biases support could look like. Targets #6. Let's discuss if you guys like it so far or if it should be done differently.

Not sure if this can be tested properly, as it requires logging in to wandb in some way.

New feature

• Weights and biases support; added to DrugGNN, MOLIR, PharmaFormer, SimpleNeuralNetwork, MultiOmicsNeuralNetwork, SuperFELTR -> new parameter --wandb_project

[nictru]

I can't comprehend why isort is throwing an error here, I tried changing the import order in the faulty file several times and it always corrects it to the version that is in this PR

[JudithBernett]

About the testing: I'd say it's okay if this functionality is not tested then. It is a nice add-on feature, but it does not change the functionality of the framework.

[nictru]

Why is it working now 🤯

[nictru]

So adding/removing the diff arg from isort seems to toggle if it errors or not

[nictru]

The first time I encountered a Heisenbug in real life (Wikipedia):

In computer programming jargon, a heisenbug is a software bug that seems to disappear or alter its behavior when one attempts to study it.

[nictru]

Okay now we can see:

 from typing import Any
 
 import numpy as np
+import wandb
 import yaml
 from sklearn.model_selection import ParameterGrid
-
-import wandb

So @JudithBernett was right, but still the local pre-commit always wants to change it to the current version

[nictru]

So it seems to work now, unfortunately my local pre-commit isort still does not match the remote CI behavior, which makes the entire pre-commit process a bit annoying

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 50.00000% with 107 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (development@f3afe32). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
drevalpy/models/drp_model.py	38.80%	41 Missing ⚠️
drevalpy/experiment.py	18.60%	35 Missing ⚠️
drevalpy/models/lightning_metrics_mixin.py	69.69%	10 Missing ⚠️
drevalpy/models/MOLIR/utils.py	36.36%	7 Missing ⚠️
drevalpy/models/SuperFELTR/utils.py	30.00%	7 Missing ⚠️
drevalpy/models/DrugGNN/drug_gnn.py	72.72%	3 Missing ⚠️
drevalpy/models/PharmaFormer/pharmaformer.py	92.00%	2 Missing ⚠️
drevalpy/models/SimpleNeuralNetwork/utils.py	77.77%	2 Missing ⚠️
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@              Coverage Diff               @@
##             development     #348   +/-   ##
==============================================
  Coverage               ?   81.63%           
==============================================
  Files                  ?       67           
  Lines                  ?     7138           
  Branches               ?        0           
==============================================
  Hits                   ?     5827           
  Misses                 ?     1311           
  Partials               ?        0           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[JudithBernett]

So it seems to work now, unfortunately my local pre-commit isort still does not match the remote CI behavior, which makes the entire pre-commit process a bit annoying

hm maybe it's the version? I have 4.5.1; the latest one

[nictru]

It's more about the isort version, I tried both 7.0.0 (latest) and 6.0.1 (used in the CI)

[PascalIversen]

Works great for me, very cool to have this! In the long run, we could make the logging better for our models with more metrics etc.

Implementation-wise, I also think it's great (apart from that it's missing from the command line interface :) )

[nictru]

I played around with wandb a bit more, tested the implementation both with hyperparameter tuning on and off to ensure the logged metrics always make sense

[JudithBernett]

looks great to me!! approved