Init_prepared_statement() got an unexpected keyword argument 'batch'

Feature Store
#1

Hello folks,

When I tried to use the online feature serving with batch, it does not work. When I use td_meta.init_prepared_statement(batch=True) it throws the following error:

image
image2088×254 41 KB

Also, get_serving_vectors does not work. I am currently using Jupyter Lab inside of Hopsworks. Since batch feature streaming looks like added to master last week, is it possible that I am using a previous version of Feature Store? How could I solve that?

Thanks.

#2

I think you might have a mismatch between the hsfs version you are running and the documentation/example you are looking at. Which version of Hopsworks/HSFS are you running?

#3

Hello Fabio,

I checked with the pip list and my hsfs version is 2.4.0, which I think is the latest version. Also, the commits of example that I am looking at it says the batch argument to init_prepared_statement and get_serving_vectors added last week to retrieve online feature serving as batch. The Jupyter notebook is here: hops-examples/feature_vector_model_serving.ipynb at 882cd104f87fd5caa153470f89c68bb6a413eca5 · logicalclocks/hops-examples · GitHub

#4

Hi - Yes 2.4.x is the latest release, unfortunately the notebook you are looking at is for the current development branch - batch retrieval from the online feature store is going to be available in the next release.

See the init_prepared_statement signature for the 2.4.x release: Training Dataset - Hopsworks Documentation

#5

Thank you! Do you know the date of next release?