chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 11 Packages Projects Releases Wiki Activity

fix: fix rmsnorm -> layernorm in qwen3 omni (#11791)

Browse Source 
Co-authored-by: Brayden Zhong <b8zhong@users.noreply.github.com>
This commit is contained in:
Vincent Zhong
2025-12-05 23:12:57 -05:00
committed by GitHub
parent a0dde90af6
commit 2ac5b98395
1 changed files with 1 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
python/sglang/srt/models/qwen3_omni_moe.py
View File

@@ -31,7 +31,6 @@ from sglang.srt.configs.qwen3_omni import (
)
from sglang.srt.configs.qwen3_vl import Qwen3VLMoeConfig
from sglang.srt.layers.attention.vision import VisionAttention
from sglang.srt.layers.layernorm import RMSNorm
from sglang.srt.layers.linear import ColumnParallelLinear, RowParallelLinear
from sglang.srt.layers.moe.fused_moe_triton.layer import FusedMoE
from sglang.srt.layers.quantization.base_config import QuantizationConfig
@@ -318,7 +317,7 @@ class Qwen3OmniMoeVisionPatchMerger(nn.Module):
super().__init__()
self.hidden_size = context_dim * (spatial_merge_size**2)
self.use_postshuffle_norm = use_postshuffle_norm
self.ln_q = RMSNorm(
self.ln_q = nn.LayerNorm(
self.hidden_size if use_postshuffle_norm else context_dim, eps=1e-6
)
self.mlp = nn.ModuleList(