topic Problem creating FeatureStore in Machine Learning

Problem creating FeatureStore

Cirsa — Thu, 03 Nov 2022 10:46:54 GMT

Hi,

When trying to create the first table in the Feature Store i get a message: ''DataFrame' object has no attribute 'isEmpty'... but it is not.

So I cannot use the function: feature_store.create_table()

With this code you should be able to reproduce the problem:

#CHUNK 1

raw_data = spark.read.format("delta").load("/databricks-datasets/nyctaxi-with-zipcodes/subsampled")

# CHUNK 2

%sql

CREATE DATABASE IF NOT EXISTS FS1;

# CHUNK 3

from databricks import feature_store

fs = feature_store.FeatureStoreClient()

fs.create_table(

name="FS1.XX",

primary_keys = ["dropoff_zip"],

df = raw_data ,

description = "Titanic Passenger Features")

AttributeError: 'DataFrame' object has no attribute 'isEmpty'

Hubert-Dudek — Thu, 03 Nov 2022 12:11:24 GMT

There is no df_train in your code. I have included the link to the feature store API doc. It can be helpful for you. Also, please set DBR to 11.3 LTS, as the API for FeatureStoreClient has changed. https://docs.databricks.com/_static/documents/feature-store-python-api-reference-0-7-0.pdf

Cirsa — Thu, 03 Nov 2022 12:38:00 GMT

@Hubert DudekSry about the 'df_train', I forgot to change it (the error I commented is real with the proper DF).

Changing the DBR to 11.3 LTS solved the problem. Thanks!

Hubert-Dudek — Thu, 03 Nov 2022 12:42:13 GMT

Great. Would you mind selecting my answer as the best one?