Add SHAP for NLP & VISION templates

[LexABzH]

Problem

We use LIME as our go to explainer for both NLP & VISION templates, but we were forced to use SHAP for the NUM template (as LIME needs the whole training dataset for each explanation).
SHAP also manage texts & images. It could be great to add it for both NLP & VISION projects.

Concerned template

• NLP template
• NUM template
• VISION template
• How templates are generated - Jinja

Solution

Add a SHAPE class to the model_explainer.py file.

Bonus : add it to our demonstrators