blog-distributed-tensorflow-map-1.py

from pyspark.sql import SparkSession
from tensorflowonspark import TFCluster, TFNode

spark = SparkSession \
        .builder \
        .config("...")
        .appName("model-training") \
        .getOrCreate()
    spark.sparkContext.addPyFile("/usr/local/tensorflow/tfspark-{version}.zip")
    
cluster = TFCluster.run(spark.sparkContext,\
                        map_fun,\
                        args, \
                        num_executors=num_executors, \
                        num_ps=num_ps,\
                        tensorboard=False,\
                        driver_ps_nodes=False,\
                        input_mode=TFCluster.InputMode.SPARK)