40 lines
1.7 KiB
Python
40 lines
1.7 KiB
Python
|
# Copyright (c) 2026 Huawei Technologies Co., Ltd. All Rights Reserved.
|
||
|
|
# Copyright 2023 The vLLM team.
|
||
|
|
#
|
||
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
||
|
|
# you may not use this file except in compliance with the License.
|
||
|
|
# You may obtain a copy of the License at
|
||
|
|
#
|
||
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
||
|
|
#
|
||
|
|
# Unless required by applicable law or agreed to in writing, software
|
||
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
||
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||
|
|
# See the License for the specific language governing permissions and
|
||
|
|
# limitations under the License.
|
||
|
|
# This file is a part of the vllm-ascend project.
|
||
|
|
from __future__ import annotations
|
||
|
|
|
||
|
|
from vllm_ascend.ops.fused_moe.moe_comm_method import AllGatherCommImpl
|
||
|
|
|
||
|
|
from .token_dispatcher import TokenDispatcherWithAllGather310
|
||
|
|
|
||
|
|
|
||
|
|
class AllGatherCommImpl310(AllGatherCommImpl):
|
||
|
|
"""This implementation is the same as NativeAllGatherCommImpl,
|
||
|
|
but uses NPU-specific ops for better performance.
|
||
|
|
|
||
|
|
This implementation should be compatible with all scenarios, and
|
||
|
|
thus it is the default implementation for MoE communication methods.
|
||
|
|
It uses `torch_npu.npu_moe_init_routing_v2` for pre-processing
|
||
|
|
and `torch_npu.npu_moe_token_unpermute` for post-processing
|
||
|
|
to handle the token-to-expert mapping and communication efficiently.
|
||
|
|
"""
|
||
|
|
|
||
|
|
def _get_token_dispatcher(self):
|
||
|
|
return TokenDispatcherWithAllGather310(
|
||
|
|
top_k=self.moe_config.experts_per_token,
|
||
|
|
num_experts=self.moe_config.num_experts,
|
||
|
|
num_local_experts=self.moe_config.num_local_experts,
|
||
|
|
)
|