[MM][Perf] Enable 2.7x faster for convolution computation with aclnn BatchMatMulV2 (#7017)
### What this PR does / why we need it? Currently, we are usinge2b31243c0/vllm/model_executor/layers/conv.py (L219-L232)for convolution computation, which is used in patch embedding for VL models. After profiling, we find that this linear method will take about **6.87 ms**, which is much slower than just using `F.conv3d()`. In `F.conv3d()`, it will call aclnn `BatchMatMulV2` with optimization on Ascend NPU, which only take about **2.50 ms** and is **2.7x faster** than linear method. - vLLM version: v0.16.0 - vLLM main:15d76f74e2--------- Signed-off-by: shen-shanshan <467638484@qq.com>
This commit is contained in:
32
vllm_ascend/ops/conv.py
Normal file
32
vllm_ascend/ops/conv.py
Normal file
@@ -0,0 +1,32 @@
|
|||||||
|
#
|
||||||
|
# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
|
||||||
|
# This file is a part of the vllm-ascend project.
|
||||||
|
#
|
||||||
|
# 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.
|
||||||
|
#
|
||||||
|
|
||||||
|
|
||||||
|
import torch
|
||||||
|
from vllm.model_executor.layers.conv import Conv2dLayer, Conv3dLayer
|
||||||
|
|
||||||
|
|
||||||
|
class AscendConv2dLayer(Conv2dLayer):
|
||||||
|
def forward_oot(self, x: torch.Tensor) -> torch.Tensor:
|
||||||
|
# Use aclnn BatchMatMulV2 for better performance on Ascend NPU.
|
||||||
|
return self._forward_conv(x)
|
||||||
|
|
||||||
|
|
||||||
|
class AscendConv3dLayer(Conv3dLayer):
|
||||||
|
def forward_oot(self, x: torch.Tensor) -> torch.Tensor:
|
||||||
|
# Use aclnn BatchMatMulV2 for better performance on Ascend NPU.
|
||||||
|
return self._forward_conv(x)
|
||||||
@@ -597,6 +597,7 @@ def register_ascend_customop(vllm_config: VllmConfig | None = None):
|
|||||||
from vllm.model_executor.custom_op import CustomOp
|
from vllm.model_executor.custom_op import CustomOp
|
||||||
|
|
||||||
from vllm_ascend.ops.activation import AscendQuickGELU, AscendSiluAndMul
|
from vllm_ascend.ops.activation import AscendQuickGELU, AscendSiluAndMul
|
||||||
|
from vllm_ascend.ops.conv import AscendConv2dLayer, AscendConv3dLayer
|
||||||
from vllm_ascend.ops.fused_moe.fused_moe import AscendFusedMoE, AscendSharedFusedMoE
|
from vllm_ascend.ops.fused_moe.fused_moe import AscendFusedMoE, AscendSharedFusedMoE
|
||||||
from vllm_ascend.ops.layernorm import AscendGemmaRMSNorm, AscendRMSNorm, AscendRMSNormGated
|
from vllm_ascend.ops.layernorm import AscendGemmaRMSNorm, AscendRMSNorm, AscendRMSNormGated
|
||||||
from vllm_ascend.ops.linear import (
|
from vllm_ascend.ops.linear import (
|
||||||
@@ -645,6 +646,8 @@ def register_ascend_customop(vllm_config: VllmConfig | None = None):
|
|||||||
"MMEncoderAttention": AscendMMEncoderAttention,
|
"MMEncoderAttention": AscendMMEncoderAttention,
|
||||||
"ApplyRotaryEmb": AscendApplyRotaryEmb,
|
"ApplyRotaryEmb": AscendApplyRotaryEmb,
|
||||||
"RMSNormGated": AscendRMSNormGated,
|
"RMSNormGated": AscendRMSNormGated,
|
||||||
|
"Conv2dLayer": AscendConv2dLayer,
|
||||||
|
"Conv3dLayer": AscendConv3dLayer,
|
||||||
}
|
}
|
||||||
|
|
||||||
# 310P: override selected ops with 310P implementations (keep minimal changes outside _310p)
|
# 310P: override selected ops with 310P implementations (keep minimal changes outside _310p)
|
||||||
|
|||||||
Reference in New Issue
Block a user