Does Data Science notebooks in R support Experiments / MLFLOW

[MatthiasA]

I'm wondering if the notebooks support mlflow using R programming language.

Similar to the free python sample notebook you get when creating a Machine Learning Model in Fabric, I've been trying to get something similar but then in R programming language.

However I continuously get errors like this when running any code with mlflow:

mlflow_set_experiment("mlflow_demo")

> Error in wait_for(function() mlflow_rest(\"experiments\", \"search\", client = client, : Operation failed after waiting for 10 seconds

There aren't really any examples (that I've found) where an experiment & model was created through MLFLOW R-API, so I'm wondering whether this is supported or not? Only examples available where models built for batch scoring.

Thanks for the feedback

[v-gchenna-msft]

Hi @MatthiasA ,

Thanks for using Fabric Community.
I would like to what challenges you are facing while working with ML Flow using R.
If you can provide some issues/challenges, I will try to address these issues with team and can you help you.

As I understand R is supported in Fabric, and mlflow is also supported in Fabric. So ideally you should be able to work with ML Flow using R.

Please let me know your issues so I can guide you better.

Docs to refer -
Machine learning experiment - Microsoft Fabric | Microsoft Learn
ML Flow in Fabric 

Hope this is helpful.

[MatthiasA]

Hello,

Thank you for your reply.

The thing is, I do not have any clue on how to create models & experiments using R.

As you mentioned, MLFlow has a python and R API, so I would expect that it is possible to create experiments & models inside a R notebook.

However, when I try to call the API with the basic function that I mentioned in my original post, I do already receive an error.

In addition, when running the Fabric data science samples 'Customer Churn' and 'Fraud detection' both in python and R, it seems that only the python scripts do create expiriments and models within Fabric. The R sample notebooks don't do this, they create models and perform a prediction, but without creating the Fabric components 'Model' and 'Experiment'.

My receiving errors when trying to create experience & model Fabric components through code in R + the Fabric samples not creating those Fabric components either, I was wondering if mlflow is actually supported (currently) for R.

All documentation that I've read about Data Science don't really specify that Fabric experiments & models are not accessible through R and I do not see anything about it on the Fabric roadmap either. But I'm unable to get this working using R.

I hope this makes it more clear for you

[v-gchenna-msft]

Hi @MatthiasA ,

Thanks for using Fabric Community.
At this time, we are reaching out to the internal team to get some help on this .
We will update you once we hear back from them.

[v-gchenna-msft · Accepted solution]

Hi @MatthiasA ,

Response from internal team -

" The mlflow integration in fabric only supports the Python models/runtime right now and R models are not supported yet. "

Unfortunately R is not supported for Python models.

[v-gchenna-msft]

Hi @MatthiasA ,

Glad to know that you got some insights over your query. Please continue using Fabric Community on your further queries.