Assign prediction to features based on manual centroid inputs in Pyspark 2.2 or 2.3

Question

I have two tables. One is a huge number of records of dense vectors of features and another is a small table of cluster number with centroid value (of the same features) in dens