I've created an ImageDataGenerator for Keras, and I would like to make sure that it runs in parallel, meaning it generates a batch while Keras processes the previous generated batch. But the parallelism is handled directly by Keras.
Is there simple test to make sure it is well parallelized?
