• This repository has been archived on 26/Dec/2018
  • Stars
    star
    298
  • Rank 139,663 (Top 3 %)
  • Language
    Python
  • Created almost 9 years ago
  • Updated about 7 years ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

TensorFlow on Spark

#Distributed TensorFlow on Spark First presented at the 2016 Spark Summit East: [Slide deck] (http://www.slideshare.net/arimoinc/distributed-tensorflow-scaling-googles-deep-learning-library-on-spark-58527889), [Presentation video] (https://www.youtube.com/watch?v=-QtcP3yRqyM), [Blog post] (https://arimo.com/machine-learning/deep-learning/2016/arimo-distributed-tensorflow-on-spark/)

##TensorSpark productionalized in yarn-cluster mode This latest version contains modifications/improvements that are mostly relevant to someone interested in taking TensorSpark to production in yarn-cluster mode (tested with a Hortonworks distribution [HDP 2.4] with CPU machines). For other deployment and machine types, the earlier version as of [Commit #62] (https://github.com/adatao/tensorspark/tree/2eae6732709884f08e800efa24653340f2f7997b) might still be a better option.

###Summary of changes since [Commit #62] (https://github.com/adatao/tensorspark/tree/2eae6732709884f08e800efa24653340f2f7997b) There are few minor improvements (see commits for details) and the following 2 major changes:

  • tensorspark.py: Reading the testset from the HDFS instead (Avoiding the need to put the testset on local disk; we are putting training and test sets at the same location on the HDFS)
  • parameterwebsocketclient.py: Find the machine that gets the Spark Driver in yarn-cluster mode (either way, there are some configs to be done here)

###To run

  1. zip pyfiles.zip ./parameterwebsocketclient.py ./parameterservermodel.py ./mnistcnn.py ./mnistdnn.py ./moleculardnn.py ./higgsdnn.py
  2. spark-submit

    --master yarn

    --deploy-mode cluster

    --queue default

    --num-executors 3

    --driver-memory 20g

    --executor-memory 60g

    --executor-cores 8

    --py-files ./pyfiles.zip

    ./tensorspark.py

Partial project layout:
tensorspark/gpu_install.sh - script to build tf from source with gpu support for aws
tensorspark/simple_websocket_*.py - simple tornado websocket example
tensorspark/parameterservermodel.py - "abstract" model class that has all tensorspark required methods implemented
tensorspark/*dnn.py - specific fully connected models for specific datasets
tensorspark/mnistcnn.py - convolutional model for mnist
tensorspark/parameterwebsocketclient.py - spark worker code
tensorspark/tensorspark.py - entry point and spark driver code