First of all, in the tensorflow documentation about Tensorboard callback https://www.tensorflow.org/api_docs/python/tf/keras/callbacks/TensorBoard , they are mentioning that you can plot histograms of activations of a model. However, when I use the tensorboard callback with model.fit, no activation is plotted. So my first Question is:
- Is there Anything Wrong? Should I be able to do it ? And before anyone asks, I'm sure I'm training the model and calling tensorboard the right way. Everything is fine, except I can't get the activation histograms.
Second Question:
- Can you tell me a way to plot the histograms of activations of a model with a custom tensorflow loop? that is, with tf.GradientTape?
Third Question:
- Can you tell me a way to plot the histograms of gradients of a model with a custom training loop? that is, with tf.GradientTape?
