Consider this:
1. The application is allowed to use only 20 GPUs.
2. To ensure exactly 20 GPUs, I use the
df.rdd.repartition(20).withResources(gpus.build).mapPartitions(func) technique.
(maxExecutors >> 20).
3. Given the volume of the input data, it takes 20 hours total to run the DL
p
I am using jupyter docker stack with spark.
So I started a new notebook and this code.
import multiprocessing
from pyspark.sql import SparkSession
from pyspark import SparkConf, SparkContext
import time
t1 = time.time()
number_cores = int(multiprocessing.cpu_count())
memory_gb = 4
def get_spar