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

Update cmdline --enable-dp-attention help string for Qwen 2/3 Moe models. (#6524)

Browse Source 
Signed-off-by: miter <miterv@outlook.com>
This commit is contained in:
miter
2025-05-24 06:20:21 +08:00
committed by GitHub
parent 9c574585b3
commit fefa19fec0
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
python/sglang/srt/server_args.py
View File

@@ -1138,7 +1138,7 @@ class ServerArgs:
parser.add_argument(
"--enable-dp-attention",
action="store_true",
help="Enabling data parallelism for attention and tensor parallelism for FFN. The dp size should be equal to the tp size. Currently only DeepSeek-V2 is supported.",
help="Enabling data parallelism for attention and tensor parallelism for FFN. The dp size should be equal to the tp size. Currently DeepSeek-V2 and Qwen 2/3 MoE models are supported.",
)
parser.add_argument(
"--enable-dp-lm-head",