a0315f6697
### What this PR does / why we need it? This pull request enables the `npugraph_ex` backend by default to improve performance on Ascend NPUs, as proposed in the [RFC](https://github.com/vllm-project/vllm-ascend/issues/6214). ### Does this PR introduce _any_ user-facing change? Yes. `npugraph_ex` is now enabled by default. Users can disable it by setting `enable: false` in the `npugraph_ex_config` section of the `additional_config`. ### How was this patch tested? CI passed. The changes are covered by existing and new E2E tests (`test_aclgraph_accuracy.py`) and unit tests (`test_ascend_config.py`) that have been updated to reflect the new default behavior. The tests verify correctness and consistency with `npugraph_ex` enabled and disabled, as well as with the new static kernel option. Signed-off-by: huyuanquan1 <huyuanquan1@huawei.com> Co-authored-by: huyuanquan1 <huyuanquan1@huawei.com>
75 lines
2.8 KiB
Python
75 lines
2.8 KiB
Python
#
|
|
# 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.
|
|
#
|
|
|
|
from torch import fx as fx
|
|
from vllm.config import VllmConfig
|
|
|
|
from vllm_ascend.ascend_config import get_ascend_config
|
|
from vllm_ascend.utils import vllm_version_is
|
|
|
|
if vllm_version_is("0.15.0"):
|
|
from vllm.compilation.inductor_pass import get_pass_context # type: ignore
|
|
from vllm.compilation.vllm_inductor_pass import VllmInductorPass # type: ignore
|
|
else:
|
|
from vllm.compilation.passes.inductor_pass import get_pass_context
|
|
from vllm.compilation.passes.vllm_inductor_pass import VllmInductorPass
|
|
|
|
|
|
class NpuGraphEXPassManager:
|
|
"""
|
|
A pass manager for npu_graph ex fusion passes.
|
|
It handles the configuration and execution of passes.
|
|
The counterpart in vllm is PostGradPassManager. Since torch_npu
|
|
does not support triton for now, we define our own pass manager.
|
|
"""
|
|
|
|
def __init__(self):
|
|
self.passes: list[VllmInductorPass] = []
|
|
|
|
def __call__(self, graph: fx.Graph) -> fx.Graph:
|
|
compile_range = get_pass_context().compile_range
|
|
|
|
for pass_ in self.passes:
|
|
if pass_.is_applicable_for_range(compile_range):
|
|
pass_(graph)
|
|
graph.recompiler()
|
|
return graph
|
|
|
|
def add(self, pass_: VllmInductorPass):
|
|
assert isinstance(pass_, VllmInductorPass)
|
|
self.passes.append(pass_)
|
|
|
|
def configure(self, config: VllmConfig):
|
|
# By default, we enable the graph fusion and quantization fusion pass.
|
|
self.npugraph_ex_config = get_ascend_config().npugraph_ex_config
|
|
|
|
if self.npugraph_ex_config.fuse_norm_quant:
|
|
from .npugraph_ex_passes.graphex_norm_quant_fusion_pass import GraphEXAddRMSNormFusionPass
|
|
|
|
self.passes.append(GraphEXAddRMSNormFusionPass(config))
|
|
|
|
if self.npugraph_ex_config.fuse_qknorm_rope:
|
|
from .npugraph_ex_passes.graphex_qknorm_rope_fusion_pass import GraphEXQKNormRopeFusionPass
|
|
|
|
self.passes.append(GraphEXQKNormRopeFusionPass(config))
|
|
|
|
if self.npugraph_ex_config.fuse_allreduce_rms:
|
|
from .npugraph_ex_passes.graphex_allreduce_rmsnorm_fusion_pass import GraphEXMatmulAllReduceAddRMSNormPass
|
|
|
|
self.passes.append(GraphEXMatmulAllReduceAddRMSNormPass(config))
|