Defining a custom model is relatively simple. First, you have to install felt pip package (this guide was created for version 0.1.0):
pip install feltoken
Then you will need to define API token for the web3.storage as an environment variable:
export WEB3_STORAGE_TOKEN='ab...'
Then you just need to pick a model from scikit-learn. For example, here we will use linear regression with L2 regularization setting the custom value of alpha. After defining the model you just need to pass it to upload_model(model) function.
Short script for defining a model and printing the CID.
Then just run the code and it will return the model CID which you can then use in web application during training plan definition.