1. Pre-requisite

In order to monitor your externally deployed model on Picsellia you have to configure:

  • A Modelon Picsellia
  • A ModelVersion related
  • Add a LabelMap for your ModelVersion
  • A Deploymentin monitoring mode with your ModelVersionattached.

1. Variables

Let's say you have a PyTorch Classification Model called Silicon-Valley-Classification with a version called Deployed predicting the classes HOT-DOG & NOT-HOT-DOG with the Labelmap:

{
  '0': 'HOT-DOG',
  '1': 'NOT-HOT-DOG'
}

2. Configuration

A. Create your Picsellia Model

B. Create your ModelVersion

๐Ÿšง

You need to have the exact same order in the Labelmap of your custom ModelVersion and the Labelmap configured on Picsellia

Now you should see a new version in your ModelVersionoverview:

It's time to deploy your ModelVersion

You should select the Monitoring only mode

And now you should be good to go: