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, 20:15 UTC
[Mics] add pcp basic support to MoE model (#31003)
Qiu committed
Dec 31, 2025 at 12:01 UTC
84f6cd741b591c780b6f5ac9be05413fd50812db
1 file changed
+8
vllm/model_executor/layers/fused_moe/config.py
+8
@@ -1027,6 +1027,10 @@ class FusedMoEConfig:
1027
def dp_size(self):
1028
return self.moe_parallel_config.dp_size
1029
1030
+ @property
1031
+ def pcp_size(self):
1032
+ return self.moe_parallel_config.pcp_size
1033
+
1034
@property
1035
def ep_size(self):
1036
return self.moe_parallel_config.ep_size
@@ -1039,6 +1043,10 @@ class FusedMoEConfig:
1043
def dp_rank(self):
1044
return self.moe_parallel_config.dp_rank
1045
1046
+ @property
1047
+ def pcp_rank(self):
1048
+ return self.moe_parallel_config.pcp_rank
1049
+
1050
@property
1051
def ep_rank(self):
1052
return self.moe_parallel_config.ep_rank