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 10, 16:30 UTC
[Bugfix] Fix torch.compile error for DP + MoE on CPU Backend (#31650)
Signed-off-by: kunzh <zhikun.wu@outlook.com>
kzwrime committed
Jan 6, 2026 at 20:06 UTC
bf0f3a4638869e3fdb2df38b61f41557acc69dfe
1 file changed
+2
-2
vllm/model_executor/layers/fused_moe/layer.py
+2
-2
@@ -1899,11 +1899,11 @@ class FusedMoE(CustomOp):
1899
)
1900
1901
post_quant_allgather = (
1902
- has_flashinfer_trtllm_fused_moe()
1903
- and self.quant_method is not None
1902
+ self.quant_method is not None
1903
and self.dp_size > 1
1904
and self.use_ep
1905
and isinstance(self.quant_method, ModelOptNvFp4FusedMoE)
1906
+ and has_flashinfer_trtllm_fused_moe()
1907
)
1908
if post_quant_allgather:
1909
hidden_states_to_dispatch, extra_tensors = (