66b0781840
### What this PR does / why we need it?
Refactor the e2e testcases.
- tests/e2e/multicard/test_weight_loader.py: Remove the unused code.
- tests/e2e/singlecard/multi-modal/test_internvl.py: Move to accuracy
test.
- tests/e2e/singlecard/test_aclgraph.py: Rename the file.
- tests/e2e/singlecard/test_embedding_aclgraph.py : Combine with
tests/e2e/singlecard/test_bge_model.py
- tests/e2e/singlecard/test_completion_with_prompt_embeds.py: Delete
eager mode and modify model to Qwen3-0.6B
- tests/e2e/singlecard/test_quantization.py: Modify model to
Qwen3-0.6B-W8A8
- tests/e2e/singlecard/test_vlm.py: Modify model to Qwen3-VL-8B
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: menogrey <1299267905@qq.com>
101 lines
2.9 KiB
Python
101 lines
2.9 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 pytest
|
|
from modelscope import snapshot_download # type: ignore[import-untyped]
|
|
|
|
from tests.e2e.conftest import HfRunner, VllmRunner
|
|
from tests.e2e.utils import check_embeddings_close
|
|
|
|
MODELS = [
|
|
"Qwen/Qwen3-Embedding-0.6B", # lasttoken
|
|
"intfloat/multilingual-e5-small" # mean_tokens
|
|
]
|
|
|
|
|
|
@pytest.mark.parametrize("model", MODELS)
|
|
def test_embed_models_correctness(model: str):
|
|
queries = ['What is the capital of China?', 'Explain gravity']
|
|
|
|
model_name = snapshot_download(model)
|
|
with VllmRunner(
|
|
model_name,
|
|
runner="pooling",
|
|
enforce_eager=False,
|
|
max_model_len=None,
|
|
cudagraph_capture_sizes=[4],
|
|
) as vllm_runner:
|
|
vllm_outputs = vllm_runner.embed(queries)
|
|
|
|
with HfRunner(
|
|
model_name,
|
|
dtype="float32",
|
|
is_sentence_transformer=True,
|
|
) as hf_runner:
|
|
hf_outputs = hf_runner.encode(queries)
|
|
|
|
check_embeddings_close(
|
|
embeddings_0_lst=hf_outputs,
|
|
embeddings_1_lst=vllm_outputs,
|
|
name_0="hf",
|
|
name_1="vllm",
|
|
tol=1e-2,
|
|
)
|
|
|
|
|
|
def test_bge_model_correctness():
|
|
queries = ['What is the capital of China?', 'Explain gravity']
|
|
|
|
model_name = snapshot_download("BAAI/bge-m3")
|
|
with VllmRunner(
|
|
model_name,
|
|
runner="pooling",
|
|
enforce_eager=False,
|
|
) as vllm_aclgraph_runner:
|
|
vllm_aclgraph_outputs = vllm_aclgraph_runner.embed(queries)
|
|
|
|
with VllmRunner(
|
|
model_name,
|
|
runner="pooling",
|
|
enforce_eager=True,
|
|
) as vllm_runner:
|
|
vllm_eager_outputs = vllm_runner.embed(queries)
|
|
|
|
with HfRunner(
|
|
model_name,
|
|
dtype="float32",
|
|
is_sentence_transformer=True,
|
|
) as hf_runner:
|
|
hf_outputs = hf_runner.encode(queries)
|
|
|
|
check_embeddings_close(
|
|
embeddings_0_lst=hf_outputs,
|
|
embeddings_1_lst=vllm_eager_outputs,
|
|
name_0="hf",
|
|
name_1="vllm",
|
|
tol=1e-2,
|
|
)
|
|
|
|
check_embeddings_close(
|
|
embeddings_0_lst=vllm_eager_outputs,
|
|
embeddings_1_lst=vllm_aclgraph_outputs,
|
|
name_0="eager",
|
|
name_1="aclgraph",
|
|
tol=1e-2,
|
|
)
|