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 8, 12:30 UTC
[Bugfix] Fix weight naming in Qwen3.5 (#34313)
Signed-off-by: Roger Wang <hey@rogerw.io>
Roger Wang committed
Feb 10, 2026 at 21:37 UTC
0b20469c627e94060d1015170b186d19de1db583
1 file changed
+1
-1
vllm/model_executor/models/qwen3_5.py
+1
-1
@@ -206,7 +206,7 @@ class Qwen3_5GatedDeltaNet(Qwen3NextGatedDeltaNet):
206
output_size=self.num_v_heads,
207
bias=False,
208
quant_config=quant_config,
209
- prefix=f"{prefix}.in_proj_ba",
209
+ prefix=f"{prefix}.in_proj_b",
210
)
211
self.in_proj_a = ColumnParallelLinear(
212
input_size=self.hidden_size,