[MoE Refactor] Remove SharedFusedMoE class (#35782)
Signed-off-by: Bill Nell <bnell@redhat.com>
This commit is contained in:
@@ -41,7 +41,7 @@ from vllm.distributed import (
|
||||
)
|
||||
from vllm.model_executor.layers.activation import SiluAndMul
|
||||
from vllm.model_executor.layers.attention import Attention
|
||||
from vllm.model_executor.layers.fused_moe import SharedFusedMoE
|
||||
from vllm.model_executor.layers.fused_moe import FusedMoE
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
from vllm.model_executor.layers.linear import (
|
||||
MergedColumnParallelLinear,
|
||||
@@ -285,7 +285,7 @@ class BailingMoE(nn.Module):
|
||||
else:
|
||||
self.shared_experts = None
|
||||
|
||||
self.experts = SharedFusedMoE(
|
||||
self.experts = FusedMoE(
|
||||
shared_experts=self.shared_experts,
|
||||
num_experts=self.num_experts,
|
||||
top_k=self.top_k,
|
||||
@@ -461,7 +461,7 @@ class BailingMoeModel(nn.Module):
|
||||
return hidden_states
|
||||
|
||||
def get_expert_mapping(self) -> list[tuple[str, str, int, str]]:
|
||||
return SharedFusedMoE.make_expert_params_mapping(
|
||||
return FusedMoE.make_expert_params_mapping(
|
||||
self,
|
||||
ckpt_gate_proj_name="gate_proj",
|
||||
ckpt_down_proj_name="down_proj",
|
||||
|
||||
Reference in New Issue
Block a user