EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

fix spell error (#1654)

Browse Source 
Fix the spell error in code

- vLLM version: v0.9.1
- vLLM main:
923147b5e8

Signed-off-by: unicorn <unicorn@unicorns-MacBook-Pro.local>
Co-authored-by: unicorn <unicorn@unicorns-MacBook-Pro.local>
This commit is contained in:
wm901115nwpu
2025-07-07 20:24:42 +08:00
committed by GitHub
parent f2a20393a2
commit f08c4f15a2
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
vllm_ascend/quantization/w8a8_dynamic.py
View File

@@ -141,8 +141,8 @@ def fused_experts_with_mc2(
local_rank = torch.distributed.get_rank(group=ep_group) local_rank = torch.distributed.get_rank(group=ep_group)
all_to_all_group_size = torch.distributed.get_world_size(ep_group) all_to_all_group_size = torch.distributed.get_world_size(ep_group)
world_szie = torch.distributed.get_world_size() world_size = torch.distributed.get_world_size()
tp_size = world_szie // all_to_all_group_size tp_size = world_size // all_to_all_group_size
tp_rank = rank % tp_size tp_rank = rank % tp_size
stage1_kwargs = { stage1_kwargs = {