Modeling Guide

Training and Deploying Model with Leonardo MLF

This is an example graph using a Leonardo Machine Learning Foundation (MLF) service. To get familiar with the service, you can run the graph and use the terminals to follow the log messages describing the training and deployment of the model.

The graph is composed of the following components:
  • trainConsumer: Consumes from S3 the .csv file with the training data.

  • testConsumer: Consumes from S3 the .csv file with the testing data.

  • DataLoader: Assigns the input port for each input of the output message attributes.

  • Training: Sends to the MLF server the data that will be used for training the model. It also takes the training script which will be used for the training. After training it downloads the model.

  • ModelRepoClient: Uploads the model to the MLF server.

  • DeploymentClient: Deploys the model to the MLF server.

  • Terminal1: Displays the logs returned by the training process.

  • Terminal2: Displays the server response as free form string.

Prerequisites

  • ModelRepoClient and DeploymentClient operator configured to access the MLF api, according the documentation of the operators.

  • Configured connection to S3 and through the trainConsumer and testConsumer operators. In addition to the existence of the csv files.

  • Training operator configured according to the operator documentation.

  • Training operator configured according to the operator documentation.

Configure and Run the Graph

  1. Fill in the configurations according to the services and the desired model.
  2. Run the graph, right-click on the Terminal operator, and select Open UI.
  3. It displays the log messages while checking the status of training process.
  4. Right-click on the Terminal2 operator and select Open UI.
  5. It displays if the deployment was successful. It can take several minutes, depending on the model.