How to load spark.mllib model without SparkContext to predict?

Question

With Spark1.6.0 MLLib, I'd build a model (like RandomForest) and save to hdfs,and then is was possible to load the randomforest model from hdfs to predict without SparkContext.Now,load the model we can use like this:

val loadModel = RandomForestModel.load(sc,modelpath)

Is there a another way to load the model without sc?Thanks!

I will load this model very frequently(not in spark streaming),so I do not want to create sparkcontext frequently — shaojie, Jun 07 '17 at 02:24

score 0 · Answer 1 · answered Oct 28 '17 at 07:21

0

We train our models using spark but for serving have our own algorithm implementation (logistic regression etc) which consumes the spark trained model output but this way don't have to instantiate spark context.

answered Oct 28 '17 at 07:21

Gaurav

41
4

How to load spark.mllib model without SparkContext to predict?

1 Answers1