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

fix bailing_moe with enable_dp_attention (#10860)

Browse Source
This commit is contained in:
GuoweiWangU
2025-09-24 17:29:32 +08:00
committed by GitHub
parent 71f24ef8f6
commit cd641a995c
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
python/sglang/srt/models/bailing_moe.py
View File

@@ -45,12 +45,12 @@ from sglang.srt.layers.dp_attention import (
get_attention_dp_size, get_attention_dp_size,
get_attention_tp_rank, get_attention_tp_rank,
get_attention_tp_size, get_attention_tp_size,
is_dp_attention_enabled,
) )
from sglang.srt.layers.layernorm import RMSNorm from sglang.srt.layers.layernorm import RMSNorm
from sglang.srt.layers.linear import ( from sglang.srt.layers.linear import (
MergedColumnParallelLinear, MergedColumnParallelLinear,
QKVParallelLinear, QKVParallelLinear,
ReplicatedLinear,
RowParallelLinear, RowParallelLinear,
) )
from sglang.srt.layers.logits_processor import LogitsProcessor from sglang.srt.layers.logits_processor import LogitsProcessor
@@ -702,7 +702,7 @@ class BailingMoEModel(nn.Module):
self.embed_dim, self.embed_dim,
quant_config=quant_config, quant_config=quant_config,
prefix=add_prefix("word_embeddings", prefix), prefix=add_prefix("word_embeddings", prefix),
use_attn_tp_group=global_server_args_dict["enable_dp_lm_head"], enable_tp=not is_dp_attention_enabled(),
) )
else: else:
self.word_embeddings = PPMissingLayer() self.word_embeddings = PPMissingLayer()