I have a well-matured pretty complex spark/scala project. I need to load models generated using sk-learn and also PyTorch and make predictions on a large dataset in batch.