88d03a783f
### What this PR does / why we need it? Refactor `vllm_ascend/ops/fused_moe` to replace scattered MoE business `**kwargs` with typed request objects and explicit stage boundaries. - Prepare, dispatch, MLP, and quant stages now have clearer ownership. - Main MoE path no longer depends on business `kwargs.get(...)` lookups. - Comm and dispatcher interfaces are request-only on the main path. - UTs can assert stage-level fields directly instead of inferring behavior indirectly. ### Does this PR introduce _any_ user-facing change? No. ### How was this patch tested? CI passed. --------- Signed-off-by: linfeng-yuan <1102311262@qq.com>
51 lines
2.0 KiB
Python
51 lines
2.0 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
|
|
|
|
import torch
|
|
|
|
from vllm_ascend.ops.fused_moe.moe_comm_method import AllGatherCommImpl
|
|
from vllm_ascend.ops.fused_moe.moe_runtime_args import MoEMlpComputeInput
|
|
|
|
from .moe_mlp import unified_apply_mlp
|
|
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 __init__(self, moe_config):
|
|
super().__init__(moe_config)
|
|
self.use_fusion_ops = False
|
|
|
|
def _apply_mlp(self, mlp_compute_input: MoEMlpComputeInput) -> torch.Tensor:
|
|
return unified_apply_mlp(mlp_compute_input=mlp_compute_input)
|
|
|
|
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,
|
|
)
|