Autoint latent bottleneck features

[jmrichardson]

Hi,

After the tabular model Autoint is fit, is it possible to obtain the bottleneck features for use in a downstream model?

[dosubot[bot]]

Hey there, @jmrichardson! I'm a bot here to help you with bugs, questions, and contributions while you wait for a human maintainer. Let me know how I can assist you with the Autoint model.

Yes, it is possible to obtain the bottleneck features from a tabular model AutoInt after it is fit for use in a downstream model. You can achieve this by using the AutoIntBackbone class, which defines a forward method that processes the input data through the network and outputs the transformed features. These features can be considered as the bottleneck features.

Here's how you can obtain them:

# Assuming `model` is an instance of AutoIntModel and `input_data` is your input tensor
bottleneck_features = model.backbone(input_data)

This will give you the transformed features from the AutoInt model, which can be used as input to another model ^[1].

To continue talking to Dosu, mention @dosu.