Extract feature vector from prediction

[dpvargas]

Is there any way to obtain the feature vector that the model extracts during the process of making a prediction with given data?

[dosubot[bot]]

You can extract feature vectors from a model in PyTorch Tabular using the DeepFeatureExtractor. Here's how you can do it:

# Pass the trained model as an argument
dt = DeepFeatureExtractor(tabular_model)
# Use the train dataframe to extract the last layer features
# The `fit` method is there only for compatibility and does not perform any operation
enc_df = dt.fit_transform(train)
# Use the extracted embeddings on a new dataframe
val_transformed = dt.transform(val)

This example demonstrates how to use the DeepFeatureExtractor to obtain feature vectors from a trained model ^[1].

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Other}