Ask HN: What tools you use for Neural Network Debugging by Visualisation?

I am tired of looking into metrics such as Accuracy, Recall, Precision, F1 score and much more only.

I am more interested in analysing what the Neural Net has learned. Visualising the gradient using techniques such as guided backpop, gradCAM etc. and projecting its outputs/features to vectors in R2 or R3 using TSNE/ PCA. I find model debugging by visualising along with Metrics Analysing helps me get more intuition into what is going wrong!

What are the the most efficient tools, githubs etc. used by you for doing the same which helps you get into crux of "what is going wrong in your model?"