[MoE Refactor] Remove SharedFusedMoE class (#35782)
Signed-off-by: Bill Nell <bnell@redhat.com>
This commit is contained in:
@@ -29,7 +29,6 @@ from vllm.model_executor.layers.fused_moe.router.fused_moe_router import (
|
||||
FusedMoERouter,
|
||||
)
|
||||
from vllm.model_executor.layers.fused_moe.router.gate_linear import GateLinear
|
||||
from vllm.model_executor.layers.fused_moe.shared_fused_moe import SharedFusedMoE
|
||||
from vllm.model_executor.layers.fused_moe.unquantized_fused_moe_method import (
|
||||
UnquantizedFusedMoEMethod,
|
||||
)
|
||||
@@ -64,7 +63,6 @@ __all__ = [
|
||||
"FusedMoEPrepareAndFinalizeModular",
|
||||
"GateLinear",
|
||||
"RoutingMethodType",
|
||||
"SharedFusedMoE",
|
||||
"activation_without_mul",
|
||||
"apply_moe_activation",
|
||||
"override_config",
|
||||
|
||||
@@ -1,25 +0,0 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
|
||||
import torch
|
||||
|
||||
from vllm.model_executor.layers.fused_moe.layer import FusedMoE
|
||||
|
||||
|
||||
# TODO(bnell): Remove this entirely
|
||||
class SharedFusedMoE(FusedMoE):
|
||||
"""
|
||||
A FusedMoE operation that also computes the results of shared experts.
|
||||
If an all2all communicator is being used the shared expert computation
|
||||
can be interleaved with the fused all2all dispatch communication step.
|
||||
"""
|
||||
|
||||
def forward(
|
||||
self,
|
||||
hidden_states: torch.Tensor,
|
||||
router_logits: torch.Tensor,
|
||||
) -> torch.Tensor:
|
||||
return super().forward(
|
||||
hidden_states=hidden_states,
|
||||
router_logits=router_logits,
|
||||
)
|
||||
Reference in New Issue
Block a user