CU-869anj8ub: Add initial commit of WWC conversion

[mart-r]

This is the working_with_cogstack replacement.

It will use medcat_den under the hood for centralised model storage.

• Convert relevant parts from working_with_cogstack
• Add workflows for scripts and notebooks
• Depend on PyPI-based medcat-den
• Test against changes on core lib
• Distribution + relevant documentation

[tomolopolis]

Task linked: CU-869anj8ub Move medcat-scripts to cogstack-nlp

[alhendrickson]

Hey can you add some detail on why some of the files were left out?

EG I see that 1_create_model and 2_train_model aren't included.

One guess is that these are both completely covered in the tutorials as well.

[alhendrickson]

Approved - just wondering if this is the compete set of files or if you should add the others. Would probably prefer to move unnecessary ones, as we can delete later, over accidentally missing some useful ones, so that we dont have to go through the other repo as much

[mart-r]

Hey can you add some detail on why some of the files were left out?

EG I see that 1_create_model and 2_train_model aren't included.

One guess is that these are both completely covered in the tutorials as well.

The creation of new models isn't really something most end users do. The idea was to only keep things that are actually relevant to the majority of the people who used working_with_cogstack. Plus, this is mostly in cogstack-ops/medcat-snomed-model-creation now.

I renamed the training to finetune_models and didn't include unsupervised training parts because - again - it's not something most users seem to be using. They just build with supervised training on top of our base models. The self-supervised training also mostly happens in cogstack-ops/medcat-snomed-model-creation now.

With that said, there is some meta cat training that I didn't port over. And I can't quite recall why that is. I'll see if I've got something documented.

[mart-r]

@alhendrickson
I added a MetaCAT example as well from WWC.

The other one doesn't really fit in the way that it's laid out.

[alhendrickson]

Approved, thanks for explaining