Dask Delayed is an approach we can use to parallelize code. It can delay the dependent function calls in task graphs and provides complete user control over parallel processes while improving performance. Its lazy computation helps us control the execution of functions. However, this differs from the execution timings of functions for parallel execution.
Let's understand the concept of Dask Delayed by looking at an example:
# Import dask delayed and compute
from dask import delayed, compute
# Create delayed function
@delayed
def cube(item):
return item ** 3
# Create delayed function
@delayed
def average(items):
return sum(items)/len(items)
# create a list
item_list = [2, 3, 4]
# Compute cube of given item list
cube_list= [cube(i) for i in item_list]
# Compute average of cube_list
computation_graph = average(cube_list)
# Compute the results
computation_graph.compute()
This results in the following output:
33.0
In the preceding example, two methods, cube and average, were annotated...
