1402cf6874
### What this PR does / why we need it?
This PR builds upon PR
https://github.com/vllm-project/vllm-ascend/pull/5011 and aims to
further enhance the npu_graph_ex_passes module. Based on prior work, we
have added graph optimization support for the add_rms_quant fused
operator in scenarios where a bias term is present—ensuring the fusion
pattern is correctly registered and matched into the computation graph.
For validation, we switched to the Qwen3-235B-A22B-W8A8 model for
QKVNormRopeWithBias and Qwen3-32B model for QKVNormRope . Benchmark
results show that, compared to the unfused baseline, enabling this
fusion pass significantly improves inference throughput for W8A8
quantized models.
For more details can refer to the
RFC:https://github.com/vllm-project/vllm-ascend/issues/4715
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
```
llm = LLM(
model=model,
tensor_parallel_size=GPUs_per_dp_rank,
enforce_eager=False,
enable_expert_parallel=enable_expert_parallel,
trust_remote_code=trust_remote_code,
gpu_memory_utilization=0.98,
max_num_batched_tokens=512,
# load_format="dummy",
max_model_len=2048,
max_num_seqs=16,
quantization="ascend",
additional_config={
"refresh": True,
"enable_npugraph_ex": True
},
compilation_config={
"cudagraph_capture_sizes": [8, 16],
"cudagraph_mode": "FULL_DECODE_ONLY",
},
)
if profile_dir:
llm.start_profile()
outputs = llm.generate(prompts, sampling_params)
if profile_dir:
llm.stop_profile()
for i, output in enumerate(outputs):
if i >= 5:
break
prompt = output.prompt
generated_text = output.outputs[0].text
print(
f"DP rank {global_dp_rank}, Prompt: {prompt!r}, "
f"Generated text: {generated_text!r}"
)
```
- vLLM version: v0.13.0
- vLLM main:
2f4e6548ef
---------
Signed-off-by: cjian <2318164299@qq.com>
52 lines
1.8 KiB
Python
52 lines
1.8 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.compilation.inductor_pass import get_pass_context
|
|
from vllm.compilation.vllm_inductor_pass import VllmInductorPass
|
|
from vllm.config import VllmConfig
|
|
|
|
|
|
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.ascend_compilation_config: dict = config.additional_config.get("ascend_compilation_config", {})
|