[Feature Request]: Add LLM template

[AgatheZ]

Description

The project template ideally needs to be able to process unstructured, structured and imaging data. There is currently support for imaging data as well as a script that leverages encoder-based transformers (BERT style) for multiclass/multilabel classification.

The repository would benefit from a script that leverages decoder-based transformers (GPT style) for various clinical tasks (information extraction, summarisation, clinical decision making etc).

Brief description of the new feature

Scripts inspired from the NLP Head and Neck workshop for information extraction.

Why do you need the new feature?

GPT style LLMs are useful when very little annotated data is available and can be used to teach smaller more runnable models.

How are you going to use this new feature?

For any NLP clinical task (IE, NER, classification etc)

Solution suggestion

Scripts inspired from the NLP Head and Neck workshop for information extraction.

Suggestion of how this could be implmented