ecd4232698
### What this PR does / why we need it?
[FlashComm2](https://gitcode.com/ascend-tribe/ascend-inference-cluster/blob/main/FlashComm/FlashComm2%E5%A4%A7%E6%A8%A1%E5%9E%8B%E6%8E%A8%E7%90%86%E4%B8%AD%E4%BB%A5%E5%AD%98%E6%8D%A2%E4%BC%A0%E7%9A%84%E9%80%9A%E4%BF%A1%E4%BC%98%E5%8C%96%E6%8A%80%E6%9C%AF.pdf)
introduces redundant storage of the o_proj matrix, which imposes
pressure on GPU memory. We propose the FlashComm2+Oshard approach by
integrating the shared linear layer feature (#2931). This approach
distributes weights layer-by-layer to each GPU and accesses the o_proj
of each layer via asynchronous broadcast operations, thereby alleviating
memory pressure while achieving nearly lossless performance compared to
the original FlashComm2. This PR implements a generalized
FlashComm2+Oshard solution.
Using following env to support flashcomm2 with oshard
```shell
export VLLM_ASCEND_FLASHCOMM2_PARALLEL_SIZE=1
--additional-config '{
"layer_sharding": ["o_proj"]
}'
```
### How was this patch tested?
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: Levi-JQ <yujinqi2@huawei.com>
Co-authored-by: Levi-JQ <yujinqi2@huawei.com>
238 lines
8.2 KiB
Python
238 lines
8.2 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
|
|
#
|
|
"""Compare the short outputs of HF and vLLM when using greedy sampling.
|
|
|
|
Run `pytest tests/test_offline_inference.py`.
|
|
"""
|
|
import os
|
|
from unittest.mock import patch
|
|
|
|
import pytest
|
|
from modelscope import snapshot_download # type: ignore
|
|
from vllm import SamplingParams
|
|
|
|
from tests.e2e.conftest import VllmRunner
|
|
from tests.e2e.model_utils import check_outputs_equal
|
|
|
|
os.environ["PYTORCH_NPU_ALLOC_CONF"] = "max_split_size_mb:256"
|
|
os.environ["VLLM_WORKER_MULTIPROC_METHOD"] = "spawn"
|
|
|
|
QWEN_DENSE_MODELS = [
|
|
"vllm-ascend/Qwen3-0.6B-W8A8",
|
|
]
|
|
|
|
QWEN_W4A8_MODELS = [
|
|
"vllm-ascend/Qwen3-1.7B-W4A8-V1",
|
|
]
|
|
|
|
DEEPSEEK_W4A8_MODELS = [
|
|
"vllm-ascend/DeepSeek-V3.1-W4A8-puring",
|
|
]
|
|
|
|
|
|
def test_deepseek_multistream_moe_tp2():
|
|
example_prompts = [
|
|
"Hello, my name is",
|
|
]
|
|
dtype = "half"
|
|
max_tokens = 5
|
|
with VllmRunner(
|
|
"vllm-ascend/DeepSeek-V3-Pruning",
|
|
dtype=dtype,
|
|
tensor_parallel_size=2,
|
|
cudagraph_capture_sizes=[1, 2, 4, 8],
|
|
distributed_executor_backend="mp",
|
|
additional_config={
|
|
"enable_multistream_moe": True,
|
|
"refresh": True,
|
|
},
|
|
) as vllm_model:
|
|
vllm_model.generate_greedy(example_prompts, max_tokens)
|
|
|
|
|
|
@pytest.mark.parametrize("model", QWEN_W4A8_MODELS)
|
|
def test_qwen3_w4a8_dynamic_tp2(model):
|
|
prompts = [
|
|
"Hello, my name is",
|
|
]
|
|
max_tokens = 5
|
|
with VllmRunner(
|
|
snapshot_download(model),
|
|
max_model_len=8192,
|
|
dtype="auto",
|
|
tensor_parallel_size=2,
|
|
cudagraph_capture_sizes=[1, 2, 4, 8],
|
|
quantization="ascend",
|
|
) as vllm_model:
|
|
vllm_model.generate_greedy(prompts, max_tokens)
|
|
|
|
|
|
def test_qwen3_moe_sp_tp2() -> None:
|
|
example_prompts = [
|
|
"Hello, my name is",
|
|
]
|
|
sampling_params = SamplingParams(max_tokens=5,
|
|
temperature=0.0,
|
|
top_k=50,
|
|
top_p=0.9)
|
|
|
|
with VllmRunner(snapshot_download("Qwen/Qwen3-30B-A3B"),
|
|
dtype="auto",
|
|
tensor_parallel_size=2,
|
|
distributed_executor_backend="mp",
|
|
compilation_config={"pass_config": {
|
|
"enable_sp": True
|
|
}},
|
|
enable_expert_parallel=True,
|
|
enforce_eager=True) as vllm_model:
|
|
vllm_model.generate(example_prompts, sampling_params)
|
|
|
|
|
|
@pytest.mark.parametrize("model", DEEPSEEK_W4A8_MODELS)
|
|
@patch.dict(os.environ, {"HCCL_BUFFSIZE": "2048"})
|
|
def test_deepseek_w4a8_accuracy_tp2(model):
|
|
prompts = [
|
|
"Hello, my name is", "The president of the United States is",
|
|
"vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs"
|
|
]
|
|
vllm_ds_w4a8_answers = [
|
|
'逍遙而至地去 accrued', '平行于我udo madreHelen', 'ysteepaolis backwards Kj'
|
|
]
|
|
sampling_params = SamplingParams(max_tokens=5, temperature=0.0)
|
|
with VllmRunner(snapshot_download(model),
|
|
dtype="auto",
|
|
tensor_parallel_size=2,
|
|
cudagraph_capture_sizes=[1, 2, 4, 8],
|
|
quantization="ascend",
|
|
enable_expert_parallel=True) as vllm_model:
|
|
vllm_quant_outputs = vllm_model.model.generate(prompts,
|
|
sampling_params)
|
|
|
|
vllm_quant_outputs_list = []
|
|
for output in vllm_quant_outputs:
|
|
vllm_quant_outputs_list.append(
|
|
([output.outputs[0].index], output.outputs[0].text))
|
|
vllm_answer_list = []
|
|
vllm_answer_list = ([([0], answer) for answer in vllm_ds_w4a8_answers])
|
|
|
|
check_outputs_equal(outputs_0_lst=vllm_answer_list,
|
|
outputs_1_lst=vllm_quant_outputs_list,
|
|
name_0="vllm_quant_outputs",
|
|
name_1="vllm_answer_outputs")
|
|
|
|
|
|
@patch.dict(os.environ, {"VLLM_ASCEND_ENABLE_FLASHCOMM1": "1"})
|
|
@patch.dict(os.environ, {"VLLM_ASCEND_FLASHCOMM2_PARALLEL_SIZE": "1"})
|
|
def test_qwen3_moe_fc2_tp2() -> None:
|
|
example_prompts = [
|
|
"Hello, my name is",
|
|
]
|
|
sampling_params = SamplingParams(max_tokens=5,
|
|
temperature=0.0,
|
|
top_k=50,
|
|
top_p=0.9)
|
|
|
|
with VllmRunner(snapshot_download("Qwen/Qwen3-30B-A3B"),
|
|
dtype="auto",
|
|
tensor_parallel_size=2,
|
|
distributed_executor_backend="mp",
|
|
enable_expert_parallel=True,
|
|
enforce_eager=True) as vllm_model:
|
|
vllm_model.generate(example_prompts, sampling_params)
|
|
|
|
|
|
@patch.dict(os.environ, {"VLLM_ASCEND_ENABLE_FLASHCOMM1": "1"})
|
|
@patch.dict(os.environ, {"VLLM_ASCEND_FLASHCOMM2_PARALLEL_SIZE": "1"})
|
|
def test_qwen3_moe_fc2_oshard_tp2() -> None:
|
|
example_prompts = [
|
|
"Hello, my name is",
|
|
]
|
|
sampling_params = SamplingParams(max_tokens=5,
|
|
temperature=0.0,
|
|
top_k=50,
|
|
top_p=0.9)
|
|
|
|
with VllmRunner(
|
|
snapshot_download("Qwen/Qwen3-30B-A3B"),
|
|
dtype="auto",
|
|
tensor_parallel_size=2,
|
|
distributed_executor_backend="mp",
|
|
enable_expert_parallel=True,
|
|
enforce_eager=
|
|
True, # TODO(Levi-JQ): support graph mode for fc2 in Qwen
|
|
additional_config={"layer_sharding": ["o_proj"]}) as vllm_model:
|
|
vllm_model.generate(example_prompts, sampling_params)
|
|
|
|
|
|
@patch.dict(os.environ, {"VLLM_ASCEND_ENABLE_FLASHCOMM1": "1"})
|
|
def test_deepseek_v2_lite_fc1_tp2() -> None:
|
|
example_prompts = [
|
|
"test" * 1001,
|
|
]
|
|
sampling_params = SamplingParams(max_tokens=5,
|
|
temperature=0.0,
|
|
top_k=50,
|
|
top_p=0.9)
|
|
with VllmRunner(snapshot_download("vllm-ascend/DeepSeek-V2-Lite-W8A8"),
|
|
dtype="auto",
|
|
tensor_parallel_size=2,
|
|
distributed_executor_backend="mp",
|
|
enable_expert_parallel=True,
|
|
enforce_eager=True,
|
|
quantization="ascend") as vllm_model:
|
|
vllm_model.generate(example_prompts, sampling_params)
|
|
|
|
|
|
@pytest.mark.parametrize("model", QWEN_DENSE_MODELS)
|
|
@patch.dict(os.environ, {"VLLM_ASCEND_ENABLE_FLASHCOMM1": "1"})
|
|
def test_qwen3_dense_fc1_tp2(model):
|
|
example_prompts = [
|
|
"Hello, my name is",
|
|
]
|
|
max_tokens = 5
|
|
|
|
with VllmRunner(
|
|
snapshot_download(model),
|
|
max_model_len=8192,
|
|
dtype="auto",
|
|
tensor_parallel_size=2,
|
|
cudagraph_capture_sizes=[1, 2, 4, 8],
|
|
quantization="ascend",
|
|
) as vllm_model:
|
|
vllm_model.generate_greedy(example_prompts, max_tokens)
|
|
|
|
|
|
@pytest.mark.parametrize("model", QWEN_DENSE_MODELS)
|
|
@patch.dict(os.environ, {"VLLM_ASCEND_ENABLE_PREFETCH_MLP": "1"})
|
|
def test_qwen3_dense_prefetch_mlp_weight_tp2(model):
|
|
example_prompts = [
|
|
"Hello, my name is",
|
|
]
|
|
max_tokens = 5
|
|
|
|
with VllmRunner(
|
|
snapshot_download(model),
|
|
max_model_len=8192,
|
|
dtype="auto",
|
|
tensor_parallel_size=2,
|
|
cudagraph_capture_sizes=[1, 2, 4, 8],
|
|
quantization="ascend",
|
|
) as vllm_model:
|
|
vllm_model.generate_greedy(example_prompts, max_tokens)
|