How to run ML Inference with Apache Beam