A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 7, 20:15 UTC
[BUGFIX][Qwen3.5] Hardcode `mlp.gate` as not quantizable (#35156)
Signed-off-by: Vadim Gimpelson <vadim.gimpelson@gmail.com>
Vadim Gimpelson committed
Feb 24, 2026 at 07:42 UTC
33a0d43c7119269aa300d4341871dd46d52575d0
1 file changed
+1
-1
vllm/model_executor/models/qwen3_next.py
+1
-1
@@ -252,7 +252,7 @@ class Qwen3NextSparseMoeBlock(nn.Module):
252
config.hidden_size,
253
config.num_experts,
254
bias=False,
255
- quant_config=quant_config,
255
+ quant_config=None,
256
prefix=f"{prefix}.gate",
257
)
258