02c26dcfc7
### What this PR does / why we need it?
[Feat] Supports Aclgraph for bge-m3
### Does this PR introduce _any_ user-facing change?
### How was this patch tested?
```
pytest -s tests/e2e/singlecard/test_embedding.py
pytest -s tests/e2e/singlecard/test_embedding_aclgraph.py
```
to start an online server with bs 10, each batch's seq length=8192, we
set --max-num-batched-tokens=8192*10 to ensure encoder is not chunked:
```
vllm serve /home/data/bge-m3 --max_model_len 1024 --served-model-name "bge-m3" --task embed --host 0.0.0.0 --port 9095 --max-num-batched-tokens 81920 --compilation-config '{"cudagraph_capture_sizes":[8192, 10240, 20480, 40960, 81920]}'
```
For bs10, each batch's seq length=8192, QPS is improved from 85 to 104,
which is a 22% improvement, lots of host bound is reduced.
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
---------
Signed-off-by: xuyexiong <xuyexiong@huawei.com>
Co-authored-by: wangyongjun <1104133197@qq.com>
56 lines
1.7 KiB
Python
56 lines
1.7 KiB
Python
#
|
|
# Copyright (c) 2025 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.
|
|
# Adapted from vllm/tests/basic_correctness/test_basic_correctness.py
|
|
#
|
|
import os
|
|
|
|
import pytest
|
|
|
|
from tests.e2e.conftest import VllmRunner
|
|
from tests.e2e.utils import check_embeddings_close
|
|
|
|
os.environ["VLLM_WORKER_MULTIPROC_METHOD"] = "spawn"
|
|
|
|
MODELS = ["BAAI/bge-m3"]
|
|
|
|
|
|
@pytest.mark.parametrize("model_name", MODELS)
|
|
def test_aclgrpah_embed_models_correctness(model_name):
|
|
queries = ['What is the capital of China?', 'Explain gravity']
|
|
|
|
with VllmRunner(
|
|
model_name,
|
|
task="embed",
|
|
enforce_eager=False,
|
|
) as vllm_aclgraph_runner:
|
|
vllm_aclgraph_outputs = vllm_aclgraph_runner.encode(queries)
|
|
|
|
with VllmRunner(
|
|
model_name,
|
|
task="embed",
|
|
enforce_eager=True,
|
|
) as vllm_runner:
|
|
vllm_outputs = vllm_runner.encode(queries)
|
|
|
|
check_embeddings_close(
|
|
embeddings_0_lst=vllm_outputs,
|
|
embeddings_1_lst=vllm_aclgraph_outputs,
|
|
name_0="hf",
|
|
name_1="vllm",
|
|
tol=1e-2,
|
|
)
|