Databricks Feature Store in MLFlow run CLI command

Alex_G
New Contributor II

Hello!

I am attempting to move some machine learning code from a databricks notebook into a mlflow git repository. I am utilizing the databricks feature store to load features that have been processed. Currently I cannot get the databricks library to import when running 'mlfow run ... -b databricks`. Is it possible to use the feature store from within mlflow run cli command if the job is being executed on the databricks backend?

Thanks!

sean_owen
Databricks Employee
Databricks Employee

Hm, what error do you get? I believe you won't be able to specify the feature store library as a dependency, as it's not externally published yet, but code that uses it should run on DB ML runtimes as it already exists there

View solution in original post