Tensorflow

What is nice about Tensorflow is that visualization is nice with tensorboard. I don’t know how you can do that with PyTorch

Keras

Keras is a really high-level API. I used to love Keras, however I always feel like I am cheating whenever I use Keras, because I don’t actually understand what is happening underneath.

• But on the other hand, why do I need to, when Keras can perfectly do what I want it to do?
• Because things like Neural Net weight initialization is super important. Are there ways to debug this?

Tensorboard

tensorboard --logdir mylogdir