Importance of input variables in neural networks?

I did a quick search for a neural network analysis tool for pytorch, but did not find much.

To start of I want to compare the importance of input variables for the test result. Later on I might want to do more, but this is what I want to start with.

Is it a function, feature, or tool - with Pytorch thank can help me find the importance of input variables of the NN?

Or is it better to use PCA, Boosted Tree algorithm, etc. - before hand - to find the importance of variables?

Thank you!

It’s not that easy to answer the question as there are many techniques to visualize the importance of the input regions and weights.
This Distill Pub publication is an awesome explanation of some concepts.