EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

[fix] PD disaggregation when enable mtp and tp!=dp (#7420)

Browse Source
This commit is contained in:
Atream
2025-06-22 03:03:11 +08:00
committed by GitHub
parent 5ea5d22170
commit 02bf31ef29
3 changed files with 4 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
python/sglang/srt/layers/dp_attention.py
View File

@@ -310,4 +310,4 @@ def attn_tp_reduce_scatter(
def attn_tp_all_gather(output_list: List[torch.Tensor], input_: torch.Tensor):
return get_attention_tp_group().all_gather(input_, tensor_list=output_list)
return get_attention_tp_group().all_gather(input_, output_tensor_list=output_list)