When training two different ML models with an increased number of workers in GCP, I get the following behavior. My question is, if we compare ML1 and ML2 in terms of time taken to train as we are increasing the number of worker nodes, what can be the reason for this behavior?
