How to parallelize a loop with Dask?

Question

I find the Dask documentation quite confusing. Let's say I have a function:

import random
import dask

def my_function(arg1, arg2, arg3):
    val = random.uniform(arg1, arg2) 
    va2 = random.uniform(arg2, arg3)
    return val1 + val2

some_list = []
for i in range(100):
    some_num = dask.delayed(my_function)(arg1, arg2, arg3)
    some_list += [some_num]

computed_list = dask.compute(*some_list)

This thing is going to fail, due to my_function() not getting all 3 arguments.

How can I parallelize this snippet of code in dask?

EDIT:

Seems to work if you put a @dask.delayed decorator on top of the function def and call it normally, but now the .compute()-method line throws:

KilledWorker: ('my_function-ac3c88f1-53f8-4d36-a520-ff8c40c6ee61', <Worker 'tcp://127.0.0.1:35925', name: 1, memory: 0, processing: 10>)

Michał Zawadzki · Accepted Answer · 2020-04-15 10:20:26Z

4

I build a graph first and then call compute on it:

import random
import dask

@dask.delayed
def my_function(arg1, arg2, arg3):
    val1 = random.uniform(arg1, arg2) 
    val2 = random.uniform(arg2, arg3)
    return val1 + val2

arg1 = 1
arg2 = 2
arg3 = 3

some_list = []
for i in range(10):
    some_num = my_function(arg1, arg2, arg3)
    some_list.append(some_num)

graph = dask.delayed()(some_list)
# graph.visualize()
computed_list = graph.compute()

answered Apr 15, 2020 at 10:20

Michał Zawadzki

7807 silver badges17 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

How to parallelize a loop with Dask?

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related