Dask backend is based on the Dask library. You can use
dask on the single machine and on the cluster thus you can simply start with the simple solution and then scale it. Additional
dask advantage is that after featching historical features you can continue with the distributed training.
To configure the polars offline store edit
project: feature_repo registry: data/registry.db provider: local online_store: ... offline_store: type: yummy.YummyOfflineStore backend: dask