I've created an open source package to visualize the forward pass any Pytorch model within notebooks like Jupyter, Google Colab and VSCode notebook. It's called torchvista.
Unlike many other tools, torchvista lets you interactively explore very large models with dragging/zooming support, and also allowing you to expand/collapse nested modules.
Also, it helps debug errors in models (it works even if the model fails midway during the forward pass).
I've created an open source package to visualize the forward pass any Pytorch model within notebooks like Jupyter, Google Colab and VSCode notebook. It's called torchvista.
Unlike many other tools, torchvista lets you interactively explore very large models with dragging/zooming support, and also allowing you to expand/collapse nested modules.
Also, it helps debug errors in models (it works even if the model fails midway during the forward pass).
There are some web-based demos in the Github of well known models if you want to check them out before installing it via pip: https://sachinhosmani.github.io/torchvista/
Keen to hear your feedback!
does it work with any model (like large transformer based models)?