tuning.py

als = ALS(maxIter = 10 ,
          userCol = "user_id", 
          itemCol = "isbn_indexed", 
          ratingCol = "book_rating",
          nonnegative = True, 
          coldStartStrategy = 'drop')

from pyspark.ml.tuning import ParamGridBuilder,CrossValidator 
grid = ParamGridBuilder().addGrid(als.rank, [10,30])\
                               .addGrid(als.regParam, [0.2,0.01,1,2])\
                               .build()

cv = CrossValidator(estimator = als,estimatorParamMaps = grid, evaluator = evaluator, parallelism = 4, numFolds = 3)

model = cv.fit(training)
best_model = model.bestModel

predictions = best_model.transform(test) 
rmse = evaluator.evaluate(predictions)
print("Root mean squared error:  ",rmse)