EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

support basic long_seq feature st (#5140)

Browse Source 
### What this PR does / why we need it?
support basic long_seq feature st 

- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c

---------

Signed-off-by: LookAround <lixushi@huawei.com>
This commit is contained in:
LookAround0301
2025-12-19 10:50:01 +08:00
committed by GitHub
parent cee9b715b5
commit 76e58d66be
2 changed files with 142 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
.github/workflows/_e2e_test.yaml vendored
View File

@@ -267,6 +267,7 @@ jobs:
pytest -sv --durations=0 tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_DeepSeek_multistream_moe
pytest -sv --durations=0 tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_Kimi_K2_Thinking_W4A16
pytest -sv --durations=0 tests/e2e/multicard/test_data_parallel_tp2.py
pytest -sv --durations=0 tests/e2e/multicard/long_sequence/test_basic.py
- name: Install Ascend toolkit & triton_ascend (for Qwen3-Next-80B-A3B-Instruct)
shell: bash -l {0}

141
tests/e2e/multicard/long_sequence/test_basic.py Normal file
View File

@@ -0,0 +1,141 @@
#
# 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
#
"""Compare the short outputs of HF and vLLM when using greedy sampling.
Run `pytest tests/e2e/multicard/test_qwen3_moe.py`.
"""
import os
import pytest
from vllm import SamplingParams
from tests.e2e.conftest import VllmRunner
from vllm_ascend.utils import vllm_version_is
os.environ["HCCL_BUFFSIZE"] = "768"
@pytest.mark.skipif(vllm_version_is('0.12.0'),
reason="0.12.0 is not supported for context sequence.")
def test_pcp_dcp_basic():
prompts = [
"The capital of France is", "Hello, my name is Tom, I am",
"The president of United States is", "AI future is"
]
model = "deepseek-ai/DeepSeek-V2-Lite-Chat"
sampling_params = SamplingParams(max_tokens=32, temperature=0.0)
with VllmRunner(model,
enforce_eager=True,
max_model_len=1024,
tensor_parallel_size=2,
prefill_context_parallel_size=2,
decode_context_parallel_size=2,
max_num_batched_tokens=1024,
enable_expert_parallel=True,
block_size=128) as runner:
runner.model.generate(prompts, sampling_params)
model = "vllm-ascend/Qwen3-30B-A3B-W8A8"
with VllmRunner(
model,
enforce_eager=True,
max_model_len=1024,
tensor_parallel_size=2,
prefill_context_parallel_size=2,
decode_context_parallel_size=1,
enable_expert_parallel=True,
block_size=128,
quantization="ascend",
) as runner:
runner.model.generate(prompts, sampling_params)
@pytest.mark.skipif(vllm_version_is('0.12.0'),
reason="0.12.0 is not supported for context sequence.")
def test_pcp_dcp_full_graph():
prompts = [
"The capital of France is", "Hello, my name is Tom, I am",
"The president of United States is", "AI future is"
]
model = "deepseek-ai/DeepSeek-V2-Lite-Chat"
sampling_params = SamplingParams(max_tokens=32, temperature=0.0)
with VllmRunner(model,
enforce_eager=False,
max_model_len=1024,
tensor_parallel_size=2,
prefill_context_parallel_size=2,
decode_context_parallel_size=2,
max_num_batched_tokens=1024,
enable_expert_parallel=True,
block_size=128,
compilation_config={
"cudagraph_mode": "FULL_DECODE_ONLY",
"cudagraph_capture_sizes": [4, 8, 24, 48, 60]
}) as runner:
runner.model.generate(prompts, sampling_params)
model = "vllm-ascend/Qwen3-30B-A3B-W8A8"
with VllmRunner(model,
enforce_eager=False,
max_model_len=1024,
tensor_parallel_size=2,
prefill_context_parallel_size=2,
decode_context_parallel_size=1,
enable_expert_parallel=True,
block_size=128,
quantization="ascend",
compilation_config={
"cudagraph_mode": "FULL_DECODE_ONLY",
"cudagraph_capture_sizes": [4, 8, 24, 48, 60]
}) as runner:
runner.model.generate(prompts, sampling_params)
@pytest.mark.skipif(vllm_version_is('0.12.0'),
reason="0.12.0 is not supported for context sequence.")
def test_pcp_dcp_piece_wise():
prompts = [
"The capital of France is", "Hello, my name is Tom, I am",
"The president of United States is", "AI future is"
]
model = "deepseek-ai/DeepSeek-V2-Lite-Chat"
sampling_params = SamplingParams(max_tokens=32, temperature=0.0)
with VllmRunner(model,
enforce_eager=False,
max_model_len=1024,
tensor_parallel_size=2,
prefill_context_parallel_size=2,
decode_context_parallel_size=2,
max_num_batched_tokens=1024,
enable_expert_parallel=True,
block_size=128) as runner:
runner.model.generate(prompts, sampling_params)
model = "vllm-ascend/Qwen3-30B-A3B-W8A8"
with VllmRunner(model,
enforce_eager=False,
max_model_len=1024,
tensor_parallel_size=2,
prefill_context_parallel_size=2,
decode_context_parallel_size=1,
enable_expert_parallel=True,
block_size=128,
quantization="ascend") as runner:
runner.model.generate(prompts, sampling_params)