8cfd257992
### What this PR does / why we need it?
Remove ETP/EP maintained in branch main. We drop this as there is no
relevant scenarios to use ETP now, and we may subsequently advocate
implementing expert tensor parallelism in vLLM to support scenarios
where the expert is needed to be sliced
This is a part of #1422 backport.
Fixes https://github.com/vllm-project/vllm-ascend/issues/1396
https://github.com/vllm-project/vllm-ascend/issues/1154
### Does this PR introduce _any_ user-facing change?
We'll not maintain etp/ep in vllm-ascend anymore, and use the tp/ep in
vllm instead.
### How was this patch tested?
CI passed with new added and existing test.
- vLLM version: v0.9.2
- vLLM main:
fe8a2c544a
Signed-off-by: MengqingCao <cmq0113@163.com>
78 lines
2.7 KiB
Python
78 lines
2.7 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.
|
|
#
|
|
"""
|
|
Execute the inference of fused_moe_allgather_ep and fused_moe_alltoall_ep.
|
|
|
|
Run 'pytest tests/multicard/test_fused_moe_allgather_ep.py'.
|
|
"""
|
|
|
|
import os
|
|
from unittest.mock import patch
|
|
|
|
from modelscope import snapshot_download # type: ignore
|
|
from vllm import SamplingParams
|
|
|
|
from tests.e2e.conftest import VllmRunner
|
|
|
|
|
|
@patch.dict(
|
|
os.environ, {
|
|
"VLLM_WORKER_MULTIPROC_METHOD": "spawn",
|
|
"TASK_QUEUE_ENABLE": "1",
|
|
"VLLM_ENABLE_FUSED_EXPERTS_ALLGATHER_EP": "1"
|
|
})
|
|
def test_generate_with_allgather():
|
|
example_prompts = ["Hello, my name is"]
|
|
sampling_params = SamplingParams(max_tokens=100, temperature=0.0)
|
|
|
|
with VllmRunner(snapshot_download("vllm-ascend/DeepSeek-V3-Pruning"),
|
|
tensor_parallel_size=4,
|
|
enforce_eager=True,
|
|
max_model_len=1024,
|
|
dtype="auto",
|
|
enable_expert_parallel=True,
|
|
additional_config={
|
|
"ascend_scheduler_config": {
|
|
"enabled": True,
|
|
"chunked_prefill_enabled": False,
|
|
},
|
|
}) as vllm_model:
|
|
vllm_model.generate(example_prompts, sampling_params)
|
|
|
|
|
|
@patch.dict(os.environ, {
|
|
"VLLM_WORKER_MULTIPROC_METHOD": "spawn",
|
|
"TASK_QUEUE_ENABLE": "1"
|
|
})
|
|
def test_generate_with_alltoall():
|
|
example_prompts = ["Hello, my name is"]
|
|
sampling_params = SamplingParams(max_tokens=100, temperature=0.0)
|
|
|
|
with VllmRunner(snapshot_download("vllm-ascend/DeepSeek-V3-Pruning"),
|
|
tensor_parallel_size=4,
|
|
enforce_eager=True,
|
|
max_model_len=1024,
|
|
dtype="auto",
|
|
enable_expert_parallel=True,
|
|
additional_config={
|
|
"ascend_scheduler_config": {
|
|
"enabled": True,
|
|
"chunked_prefill_enabled": False,
|
|
},
|
|
}) as vllm_model:
|
|
vllm_model.generate(example_prompts, sampling_params)
|