Building a real-time scoring service

For Azure Machine Learning, you can't really choose a specific deployment case to match your use case. To implement a real-time scoring service, you need to pick a highly scalable compute target (for example, AKS) and provide a scoring file that receives data with each request and returns the prediction of the model synchronously:

1. To do so, you need to provide the init() and run() functions in the scoring file. Let's take a look at a simple scoring file. In reality, this should be very simple, as we have seen most of the code already:

   import json
   import numpy as np
    import os
   from sklearn.externals import joblib
   def init():
       global model
       model_path = Model.get_model_path('sklearn_mnist')
       model= joblib.load(model_path)
   def run(data): try:
       result = model.predict(data)
       # You can return any JSON serializable...