What do you want to see?

[johnmarktaylor91]

Thank you for checking out torchlens! As it is still in its early days, I would love to get people's feedback for what they would find useful from it. Generally speaking, the aim of torchlens is to easily extract activations and understand a model's computational graph via visualizations and helpful metadata about a network and its layers. I am unlikely to add functionality for further analyzing the extracted activations, since there already exist excellent packages (e.g., rsatoolbox) for this, or for loading particular models (since torchlens is meant to work for any model), but beyond that the door is open for what folks would find useful from it.