calculation-graph and neural-network

[animesh]

🚀 Feature

I am wondering if there is a way to plot the calculation-graph/neural-network itself? S

Motivation

The layer visualization can make the understanding of what is going on incredibly interesting

Pitch

Apart from making it interesting, it also brings a clarity of how the data is being processed, something like https://colah.github.io/posts/2015-08-Backprop/

Alternatives

Tensorboard already has the calculation-graph interactively displayed at tensorboard-/#graphs&run=. For example when i run the code at https://github.com/animesh/dlomix/blob/develop/checkDLomix.py and change the line-28 to collect logs in tensorboard

import tensorflow as tf
tensorboard_callback = tf.keras.callbacks.TensorBoard(log_dir="/mnt/promec-ns9036k/.tools/deep-learn-1643983488-tensorboard")
history = model.fit(rtdata.train_data,validation_data=rtdata.val_data,epochs=10, callbacks=[tensorboard_callback])

I get the graph

probably directly incorporating this in aim can be a good start? next step to be actually plot the neural-network 💯

[simonrouse9461]

I found a great open-source neural net visualization project: https://github.com/lutzroeder/netron Is it possible to integrate? I don't think it's a good idea to reinvent the wheel.