xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-06-06 17:42:16 +08:00
Code Issues Packages Projects Releases Wiki Activity

minor

Browse Source 
Signed-off-by: Woosuk Kwon <woosuk@thinkingmachines.ai>
This commit is contained in:
Woosuk Kwon 2025-09-19 07:11:37 +00:00
parent 37478c18cf
commit 9c75d896a8
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
vllm/v1/worker/gpu/model_runner.py
View File

@ -381,7 +381,7 @@ class GPUModelRunner:
tp_group = get_tp_group() tp_group = get_tp_group()
tp_size = tp_group.world_size tp_size = tp_group.world_size
n = (num_reqs + tp_size - 1) // tp_size n = (num_reqs + tp_size - 1) // tp_size
use_dp_sampler = tp_size > 1 and n > 32 use_dp_sampler = tp_size > 1 and n > 32 # TODO(woosuk): Tune.
if use_dp_sampler: if use_dp_sampler:
# NOTE(woosuk): Make sure that no rank gets zero requests. # NOTE(woosuk): Make sure that no rank gets zero requests.
tp_rank = tp_group.rank tp_rank = tp_group.rank