microsoft/mscclpp
1
0
Fork 0
mirror of https://github.com/microsoft/mscclpp.git synced 2026-05-12 01:10:22 +00:00
Code Issues Packages Projects Releases Wiki Activity

review: fix docstring, trailing comma, import placement, and filename mismatch

Browse Source 
Agent-Logs-Url: https://github.com/microsoft/mscclpp/sessions/f587a2e1-568f-4596-bb02-342c101dd539

Co-authored-by: Binyang2014 <9415966+Binyang2014@users.noreply.github.com>
This commit is contained in:
copilot-swe-agent[bot]
2026-05-03 22:51:44 +00:00
committed by GitHub
parent 2a34a7ed11
commit b30752a94f
3 changed files with 9 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
python/mscclpp/__main__.py
View File

@@ -73,7 +73,7 @@ default_algo_configs = [
tags={"default": 1},
),
"additional_kwargs": {"thread_block_group_size": 8},
}
},
]

11
python/mscclpp/default_algos/allreduce_multi_nodes.py
View File

@@ -2,12 +2,13 @@
# Licensed under the MIT License.
"""
Multi-node AllReduce implementation using packet-based communication.
This implements a hierarchical AllReduce: intra-node allreduce followed by
inter-node exchange and final intra-node allreduce.
Generalized multi-node AllReduce implementation using packet-based communication.
This implements a hierarchical AllReduce for N nodes:
1. Intra-node reduce-scatter (each GPU reduces its assigned chunk across the node)
2. Inter-node allreduce (exchange fully intra-reduced chunks across all nodes)
3. Intra-node broadcast (distribute the fully reduced chunks back to all GPUs in the node)
"""
import argparse
from mscclpp.language.utils import AlgoSpec
from mscclpp.language.channel import *
from mscclpp.language.rank import *
@@ -208,6 +209,8 @@ def allreduce_multi_nodes(spec: AlgoSpec, thread_block_group_size: int) -> Colle
if __name__ == "__main__":
import argparse
parser = argparse.ArgumentParser()
parser.add_argument("--name", type=str, help="name of the program")
parser.add_argument("--num_gpus", type=int, help="total number of gpus")