6c020883a8
<!-- Thanks for sending a pull request! BEFORE SUBMITTING, PLEASE READ https://docs.vllm.ai/en/latest/contributing/overview.html --> ### What this PR does / why we need it? This PR add new function of : aclgraph_batch_size can dynamic adjust to different model; before this PR, the aclgraph_batch_sizes given from vllm to vllm-ascend always too large, and that may result in ERROR while running on different, with the information: "The resources are insufficient". Now, with this PR, the code can dynamic adjust aclgraph_batch_sizes depend on the model hidden_layer_nums and parallel config, for example: a. for Qwen2.5-7B, the aclgraph_batch_size length is 33 total; b. for Qwen2.5-72B, the aclgraph_batch_size length is 11 total; Signed-off-by: chris668899 <15105191595@126.com>
61 lines
1.9 KiB
Python
61 lines
1.9 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.
|
|
#
|
|
import pytest
|
|
import torch
|
|
from vllm import LLM, SamplingParams
|
|
|
|
# TODO: revert me when cuda hard code is fixed in 'VllmBackend'
|
|
torch.cuda.CUDAGraph = torch.npu.NPUGraph
|
|
|
|
MODELS = [
|
|
"Qwen/Qwen2.5-0.5B-Instruct",
|
|
]
|
|
|
|
TENSOR_PARALLELS = [2]
|
|
|
|
prompts = [
|
|
"Hello, my name is",
|
|
"The future of AI is",
|
|
]
|
|
|
|
|
|
@pytest.mark.parametrize("model", MODELS)
|
|
@pytest.mark.parametrize("tp_size", TENSOR_PARALLELS)
|
|
@pytest.mark.parametrize("max_tokens", [64])
|
|
@pytest.mark.parametrize("temperature", [0.0])
|
|
@pytest.mark.parametrize("ignore_eos", [True])
|
|
def test_models(model: str, tp_size: int, max_tokens: int, temperature: int,
|
|
ignore_eos: bool) -> None:
|
|
# Create an LLM.
|
|
llm = LLM(
|
|
model=model,
|
|
tensor_parallel_size=tp_size,
|
|
)
|
|
# Prepare sampling_parames
|
|
sampling_params = SamplingParams(
|
|
max_tokens=max_tokens,
|
|
temperature=temperature,
|
|
ignore_eos=ignore_eos,
|
|
)
|
|
|
|
# Generate texts from the prompts.
|
|
# The output is a list of RequestOutput objects
|
|
outputs = llm.generate(prompts, sampling_params)
|
|
torch.npu.synchronize()
|
|
# The output length should be equal to prompts length.
|
|
assert len(outputs) == len(prompts)
|