* allow custom model

[aaarrti]

Description

• Right now, user can only use TensorFlowModel or PyTorchModel from quantus
• If user wants to evaluate, e.g, baseline (sklearn) model, there is no way

Implemented changes

• Allow custom implementers of ModelInterface

[codecov-commenter]

Codecov Report

Attention: 1 lines in your changes are missing coverage. Please review.

Comparison is base (c32da2b) 91.20% compared to head (1a7d0fb) 91.18%.

Files	Patch %	Lines
quantus/helpers/utils.py	50.00%	1 Missing ⚠️

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #322      +/-   ##
==========================================
- Coverage   91.20%   91.18%   -0.03%     
==========================================
  Files          66       66              
  Lines        3887     3889       +2     
==========================================
+ Hits         3545     3546       +1     
- Misses        342      343       +1     

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

[annahedstroem]

Could there be any functions calling this function where this statement needs to hold true, i.e., where we want to return the ModelInterface from this function? @aaarrti

[aaarrti]

Could there be any functions calling this function where this statement needs to hold true, i.e., where we want to return the ModelInterface from this function? @aaarrti

Could there be any functions calling this function where this statement needs to hold true, i.e., where we want to return the ModelInterface from this function? @aaarrti

Hi,
the use case for this is custom implementations of ModelInterface provided by user, smth like:

import quantus

class MyOwnCustomModel(quantus.ModelInterface):

   """
   User-defined wrapper around model, 
   which is not supported out of the box by Quantus
   """

   def predict(x_batch, **kwargs):
       ...


model = MyOwnCustomModel()
scores = quantus.RandomLogit()(model, ...)       

[annahedstroem]

hi @aaarrti great example! can you formulate that as a test case? that would be golden!

[aaarrti]

Hi @annahedstroem,
this tiny PR was meant as a demonstration for @abarbosa94 on how we could enable custom user defined models. Unfortunately, I barely have any free time lately, so I will not be able to fully develop tests case and demo notebook showcasing custom models support. As this is a tiny change, I think it can easily be made when it is actually needed. So, I will close the PR 🙃