Training using tf.GradientTape and MirroredStrategy with Dataset.from_generator

Question

I\'m trying to combine tf.data.Dataset.from_generator and tf.GradientTape to train the network using two GPUs.

I found the example from the