Cannot Create Real-time Endpoint

Question

I tried to use the tutorial on deploying a Machine Learning Endpoint using this link: https://learn.microsoft.com/en-us/azure/machine-learning/tutorial-designer-automobile-price-deploy

First off, I created my own model using my own dataset and was able to create a model with a 96% confidence level. I then started to deploy it using the tutorial above and found two issues:

1. Upon creating the real-time inference pipeline the Web Service Input is not generated. Hence, I needed to manually add it..
2. After adding the Web Service Input and running it again, I tried to deploy it but got this error:
   Realtime endpoint deploy 'test' failed Deploy: Failed on Preparing to deploy. Details: Call MT PrepareCreateRealTimeEndpointRequest api failed. Extract schema failed for web service input input1, from data node of node id: 8c29f71d

Is there any thing I am missing out here?

Answer 1

Ziggy Zulueta Is crop_data dataset using a tabular dataset or a dataset which expects zip files? In most probability this error means that web service input did not get the same data schema as crop_data your dataset. When you connect the web input port to the input port of select columns from dataset it expects the data node i.e crop_data component to provide the schema required to build the expected schema for the endpoint input. If it is unable to do so re-check if the data node is tabular schema instead of a zip file or files.

Another way to ensure it follows or extracts the data schema is to use "Enter data manually" module and add some sample data to its settings so that the web service input component can use this schema. In this case you can safely remove the crop_data node and just add web service input and enter data manually module. Please see this page for guidance on using web service input module.