0005479b9c
### What this PR does / why we need it?
This PR prefetchs the weight of mlp layers in Qwen Dense Models to
optimize the performance in Decode phase mainly.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
CI passed with new added/existing test.
- vLLM version: main
- vLLM main:
a1213fae5f
Signed-off-by: rjg-lyh <1318825571@qq.com>
Co-authored-by: Shuming19 <313093131@qq.com>
38 lines
1.3 KiB
Python
38 lines
1.3 KiB
Python
#
|
|
# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
|
|
#
|
|
# 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.
|
|
#
|
|
|
|
import torch
|
|
|
|
|
|
def torchair_silu_and_mul_forward_oot(self, x: torch.Tensor) -> torch.Tensor:
|
|
"""AscendSiluAndMul forward in torchair mode.
|
|
|
|
The key difference from the original implementation is the removal of operators
|
|
from the torch.ops.vllm class, as these operators only function in non-torchair
|
|
modes. Adding them back would cause the graph compilation to fail.
|
|
"""
|
|
|
|
import torch_npu
|
|
|
|
from vllm_ascend.utils import is_310p
|
|
|
|
if is_310p():
|
|
out = torch_npu.npu_swiglu(x.to(torch.float32)).to(torch.float16)
|
|
else:
|
|
out = torch_npu.npu_swiglu(x)
|
|
return out
|