Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[Core][Distributed] use cpu group to broadcast metadata in cpu #4444

Merged
merged 11 commits into from
Apr 29, 2024
Merged

[Core][Distributed] use cpu group to broadcast metadata in cpu #4444

Changes from 2 commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
7649677
use cpu group to broadcast metadata in cpu
youkaichao Apr 29, 2024
93cfc0c
fix lint
youkaichao Apr 29, 2024
98537bf
add comment
youkaichao Apr 29, 2024
7f08b26
update outdated mock
youkaichao Apr 29, 2024
e674b6f
fix test
youkaichao Apr 29, 2024
2c517a4
fix lint
youkaichao Apr 29, 2024
8c5f0ca
fix local rank
youkaichao Apr 29, 2024
5c526e1
use kwargs
youkaichao Apr 29, 2024
4e49c09
refactor args
youkaichao Apr 29, 2024
91dbbe7
add _split_tensor_dict
youkaichao Apr 29, 2024
36fb2ea
Merge remote-tracking branch 'origin' into broadcast_cpu
youkaichao Apr 29, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 9 additions & 3 deletions vllm/distributed/communication_op.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,8 @@
import torch
from torch.distributed import ProcessGroup

from .parallel_state import (get_tensor_model_parallel_group,
from .parallel_state import (get_cpu_world_group,
get_tensor_model_parallel_group,
get_tensor_model_parallel_rank,
get_tensor_model_parallel_world_size,
is_pynccl_enabled_for_all_reduce)
Expand Down Expand Up @@ -146,6 +147,11 @@ def broadcast_tensor_dict(
group: Optional[ProcessGroup] = None,
) -> Optional[Dict[Any, Union[torch.Tensor, Any]]]:
"""Broadcast the input tensor dictionary."""
if group is None:
group = torch.distributed.group.WORLD
cpu_group = get_cpu_world_group()
else:
cpu_group = group
youkaichao marked this conversation as resolved.
Show resolved Hide resolved
group = group or torch.distributed.group.WORLD
ranks = torch.distributed.get_process_group_ranks(group)
assert src in ranks, f"Invalid src rank ({src})"
Expand All @@ -172,7 +178,7 @@ def broadcast_tensor_dict(
metadata_list.append((key, value))
torch.distributed.broadcast_object_list([metadata_list],
src=src,
group=group)
group=cpu_group)
async_handles = []
for key, value in metadata_list:
if isinstance(value, TensorMetadata):
Expand All @@ -189,7 +195,7 @@ def broadcast_tensor_dict(
recv_metadata_list = [None]
torch.distributed.broadcast_object_list(recv_metadata_list,
src=src,
group=group)
group=cpu_group)
assert recv_metadata_list[0] is not None
tensor_dict = {}
async_handles = []
Expand Down
Loading