PySpark regression models giving different accuracy results on a single node setup vs multi node cluster

Question

The below code shows how I am creating the model pipeline, in this only the regressor changes. I am trying with RF, LR, and GBR models but all of them show different results whe