Icon

02_​Prediction_​workflow_​optimization

There has been no title set for this workflow's metadata.

Prediction workflow. In this workflow we optimize the prediction workflow directly. The same can be done from the training workflow if the prediction workflow is created from the training workflow automatically. The prediction workflow is optimized by adding streaming to the preprocessing to speed up the inference.

Step 1. Preprocess input data in chunks

  1. Convert the Data Preprocessing (Apply) metanode to a component

  2. In the Job Manager Selection tab of the component configuration, select Simple Streaming with chunk size 50*


* You need the KNIME Streaming Execution (Beta) extension



Part 4 - Best Practices when Productionizing Data Science

Exercise 02 Prediction workflow optimization

Learning objective: In this exercise you'll learn how to optimize the prediction workflow.


Workflows description: Prediction workflow. In this workflow we optimize the prediction workflow directly. The same can be done from the training workflow if the prediction workflow is created from the training workflow automatically. The prediction workflow is optimized by adding streaming to the preprocessing to speed up the inference.


You'll find the instructions to the exercises in the yellow annotations.

Container Input (Table)
PortObject Reference Reader
Readers
XGBoost Predictor
Container Output (Table)
Data Preprocessing (Apply)

Nodes

Extensions

Links