biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

Re-enable the 80 char line width limit (#3305)

Browse Source
This commit is contained in:
Zhuohan Li
2024-03-10 19:49:14 -07:00
committed by GitHub
parent 4b59f00e91
commit 2f8844ba08
67 changed files with 557 additions and 528 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
vllm/model_executor/parallel_utils/communication_op.py
View File

@@ -11,7 +11,8 @@ from vllm.model_executor.parallel_utils.parallel_state import (
get_tensor_model_parallel_group,
is_cupy_nccl_enabled_for_all_reduce,
)
from vllm.model_executor.parallel_utils.custom_all_reduce import custom_all_reduce
from vllm.model_executor.parallel_utils.custom_all_reduce import (
custom_all_reduce)
def tensor_model_parallel_all_reduce(input_: torch.Tensor) -> torch.Tensor:
@@ -24,7 +25,7 @@ def tensor_model_parallel_all_reduce(input_: torch.Tensor) -> torch.Tensor:
and GPU topology.
TLDR: always assume this function modifies its input, but use the return
value as the output.
value as the output.
"""
# Bypass the function if we are using only 1 GPU.
if get_tensor_model_parallel_world_size() == 1: