Exercise 6 - train_step/train_step function lines 29 and 31
grad = tape.gradient(J, generated_image)
optimizer.apply_gradients([(grad, generated_image)])
why do we pass both grad which contains pairs of cost function and generated image and again generated image when we use “apply_gradient” method above?
from Tensorflow documentation, the arguments are
grads_and_vars List of (gradient, variable) pairs as returned by compute_gradients().
global_step Optional Variable to increment by one after the variables have been updated.
name Optional name for the returned operation. Default to the name passed to the Optimizer constructor.
in the lab code, does the [(grad, generated_image)]) correspond to “grads_and_vars” argument? if so, why do we repeat generated_image? or is generated image correspond to a different arg?
https://www.tensorflow.org/api_docs/python/tf/compat/v1/train/Optimizer#apply_gradients
