Created
June 15, 2022 16:58
-
-
Save stephanie-wang/898964c7f37b3f872bcfcaba1ab603d5 to your computer and use it in GitHub Desktop.
map reduce with generators
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import ray | |
@ray.remote | |
def map(start, end, boundaries): | |
vals = list(range(start, end)) | |
partitions = [] | |
prev_bound = 0 | |
for next_bound in boundaries: | |
partitions.append([x for x in vals if x >= prev_bound and x < next_bound]) | |
partitions.append([x for x in vals if x >= prev_bound]) | |
return partitions | |
@ray.remote | |
def reduce(*map_results): | |
map_results = [x for y in map_results for x in y] | |
return sorted(map_results) | |
@ray.remote | |
def reduce_generator(multiple_map_results): | |
for map_results in multiple_map_results: | |
# NOTE: Now we're passing a doubly nested list of ObjectRefs and we | |
# don't pass them as direct args, so we have to call ray.get here to | |
# get the actual values. | |
map_results = ray.get(map_results) | |
map_results = [x for y in map_results for x in y] | |
yield sorted(map_results) | |
if __name__ == "__main__": | |
# 1 reduce output per task. | |
num_map = 10 | |
num_reduce = 2 | |
boundaries = [50] | |
map_results = [ | |
map.options(num_returns=len(boundaries) + 1).remote(0, 100, boundaries) | |
for _ in range(num_map)] | |
reduce_results = [] | |
for i in range(num_reduce): | |
reduce_results.append(reduce.remote(*[map_result[i] for map_result in map_results])) | |
# Same number of reduce tasks, but multiple outputs per task. | |
num_map = 10 | |
num_reduce = 10 | |
num_reduce_tasks = 2 | |
boundaries = [10, 20, 30, 40, 50, 60, 70, 80, 90] | |
map_results = [map.options(num_returns=len(boundaries) + 1).remote(0, 100, boundaries) | |
for _ in range(num_map)] | |
reduce_results = [] | |
num_reduce_returns = num_reduce // num_reduce_tasks | |
for i in range(num_reduce_tasks): | |
reduce_args = [] | |
for j in range(num_reduce_returns): | |
reduce_args.append([map_result[i * num_reduce_returns + j] for map_result in map_results]) | |
reduce_results += reduce_generator.options(num_returns=num_reduce_returns).remote(reduce_args) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment