Classifier from feature extractor

[code-demoe]

How can I create a classifier network from a feature extractor?
Given the output of a feature extractor is a list of tensors, to create a classifier I will need to pass in each tensor to the model at different layers. But the names returned by model.feature_info.module_name() corresponds to the full model instead of the feature extractor, so finding the layers becomes a bit tricky. How to approach this problem?