Remove MLFlow Integration from Base Environment for Streamlined Configuration

[lvijnck]

Description

to add

How Has This Been Tested?

Please describe the tests that you ran to verify your changes. Provide instructions so we can reproduce. Please also list any relevant details for your test configuration

• Re-ran against prod

Checklist:

• added label to PR (e.g. enhancement or bug)
• I have looked at the diff on github to make sure no unwanted files have been committed.
• I have made corresponding changes to the documentation
• I have added tests that prove my fix is effective or that my feature works
• Any dependent changes have been merged and published in downstream modules
• ran /run-tests check at the end of PR collaboration work to execute integration tests

[piotrkan]

Why are we removing MLFlow from base @lvijnck ? It's useful even locally, I used it myself quite a lot when doing embedding benchmark work

[lvijnck]

Why are we removing MLFlow from base @lvijnck ? It's useful even locally, I used it myself quite a lot when doing embedding benchmark work

What exactly do you use locally in that case? For lot's of use-cases it's really in the way, as it adds a stateful component to local testing.

That being said, I think we need to go for the ability to run locally either with MLFlow, or without MLFlow. Though Kedro does not support that currently.

[piotrkan]

I used it mainly for comparing PCA plots and convergence plots between different Node2vec/Graphsage runs for the subsamples and once we have a sample of the data as mentioned in #348 we can also run modelling tests and compare how the modelling runs compare. Also having it locally can be good for development work, e.g. for inference pipeline I used it for extracting the model from local mlflow artifact storage.

If it was up to me whether we should keep it in base, I would keep it but maybe let's discuss it in a broader group. I agree that ideal scenario would be to have a choice to run the pipeline with or without mlflow tracking