Predicting in parallel using concurrent.futures of tensorflow.keras models

Question

I am trying to implement some parallel jobs using concurrent.futures. Each worker requires a copy of a TensorFlow model and some data. I implemented it in the follo