Maybe load the model on each executor’s disk and load it from there? Depending on how you use the data/model, using something like Livy and sharing the same connection may help?

 

From: Naveen Swamy [mailto:mnnaveen@gmail.com]
Sent: Wednesday, September 27, 2017 9:08 PM
To: user@spark.apache.org
Subject: Loading objects only once

 

Hello all,

 

I am a new user to Spark, please bear with me if this has been discussed earlier. 

 

I am trying to run batch inference using DL frameworks pre-trained models and Spark. Basically, I want to download a model(which is usually ~500 MB) onto the workers and load the model and run inference on images fetched from the source like S3 something like this

rdd = sc.parallelize(load_from_s3)

rdd.map(fetch_from_s3).map(read_file).map(predict)

 

I was able to get it running in local mode on Jupyter, However, I would like to load the model only once and not every map operation. A setup hook would have nice which loads the model once into the JVM, I came across this JIRA https://issues.apache.org/jira/browse/SPARK-650  which suggests that I can use Singleton and static initialization. I tried to do this using a Singleton metaclass following the thread here https://stackoverflow.com/questions/6760685/creating-a-singleton-in-python. Following this failed miserably complaining that Spark cannot serialize ctype objects with pointer references.

 

After a lot of trial and error, I moved the code to a separate file by creating a static method for predict that checks if a class variable is set or not and loads the model if not set. This approach does not sound thread safe to me, So I wanted to reach out and see if there are established patterns on how to achieve something like this.

 

 

Also, I would like to understand the executor->tasks->python process mapping, Does each task gets mapped to a separate python process?  The reason I ask is I want to be to use mapPartition method to load a batch of files and run inference on them separately for which I need to load the object once per task. Any 

 

 

Thanks for your time in answering my question.

 

Cheers, Naveen