Skip to content

Commit 5347f2d

Browse files
jesse996jesse996
committed
Merge branch 'main' into event-sync
Signed-off-by: jesse <[email protected]>
2 parents dd4c177 + cd58a64 commit 5347f2d

File tree

399 files changed

+47971
-10561
lines changed

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

399 files changed

+47971
-10561
lines changed

‎.github/Dockerfile.buildwheel‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -15,7 +15,7 @@
1515
# This file is a part of the vllm-ascend project.
1616
#
1717
ARG PY_VERSION=3.11
18-
FROM quay.io/ascend/manylinux:8.0.0-910b-manylinux_2_28-py${PY_VERSION}
18+
FROM quay.io/ascend/manylinux:8.2.rc1-910b-manylinux_2_28-py${PY_VERSION}
1919

2020
ARG COMPILE_CUSTOM_KERNELS=1
2121

‎.github/ISSUE_TEMPLATE/900-release-checklist.yml‎

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -24,6 +24,8 @@ body:
2424
value: >
2525
- [ ] Create a new issue for release feedback
2626
27+
- [ ] Upgrade vllm version to the new version for CI and Dockerfile
28+
2729
- [ ] Write the release note PR.
2830
2931
- [ ] Update the feedback issue link in docs/source/faqs.md

‎.github/actionlint.yaml‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -18,3 +18,4 @@ self-hosted-runner:
1818
- linux-amd64-cpu-0
1919
- linux-amd64-cpu-8
2020
- linux-amd64-cpu-16
21+
- linux-aarch64-a3-0

‎.github/workflows/_accuracy_test.yaml‎

Lines changed: 175 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,175 @@
1+
name: 'accuracy test'
2+
3+
on:
4+
workflow_call:
5+
inputs:
6+
vllm:
7+
required: true
8+
type: string
9+
vllm-ascend:
10+
required: false
11+
type: string
12+
default: main
13+
runner:
14+
required: true
15+
type: string
16+
image:
17+
required: true
18+
type: string
19+
model_name:
20+
required: true
21+
type: string
22+
upload:
23+
required: false
24+
type: boolean
25+
default: false
26+
27+
jobs:
28+
accuracy_tests:
29+
30+
runs-on: ${{ inputs.runner }}
31+
name: ${{ inputs.model_name }} accuracy
32+
container:
33+
image: swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/cann:8.2.rc1-910b-ubuntu22.04-py3.11
34+
env:
35+
VLLM_USE_MODELSCOPE: True
36+
# 1. If version specified (work_dispatch), do specified branch accuracy test
37+
# 2. If no version (labeled PR), do accuracy test by default ref:
38+
# The branch, tag or SHA to checkout. When checking out the repository that
39+
# triggered a workflow, this defaults to the reference or SHA for that event.
40+
# Otherwise, uses the default branch.
41+
GHA_VLLM_ASCEND_VERSION: ${{ inputs.vllm-ascend }}
42+
43+
steps:
44+
- name: Checkout repository
45+
uses: actions/checkout@v4
46+
47+
- name: Set model name as output
48+
id: set_output
49+
run: |
50+
echo "model_name=${{ inputs.model_name }}" >> $GITHUB_OUTPUT
51+
52+
- name: Config mirrors
53+
run: |
54+
sed -Ei 's@(ports|archive)[email protected]:8081@g' /etc/apt/sources.list
55+
pip config set global.index-url http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple
56+
pip config set global.trusted-host cache-service.nginx-pypi-cache.svc.cluster.local
57+
apt-get update -y
58+
apt install git -y
59+
60+
- name: Install system dependencies
61+
run: |
62+
apt-get -y install `cat packages.txt`
63+
apt-get -y install gcc g++ cmake libnuma-dev
64+
65+
- name: Checkout vllm-project/vllm repo
66+
uses: actions/checkout@v4
67+
with:
68+
repository: vllm-project/vllm
69+
ref: ${{ inputs.vllm }}
70+
path: ./vllm-empty
71+
72+
- name: Install vllm-project/vllm from source
73+
working-directory: ./vllm-empty
74+
run: |
75+
VLLM_TARGET_DEVICE=empty pip install -e .
76+
77+
- name: Resolve vllm-ascend version
78+
run: |
79+
VERSION_INPUT="${{ inputs.vllm-ascend }}"
80+
81+
if [[ "$VERSION_INPUT" == "latest" ]]; then
82+
TAGS=$(git ls-remote --tags --sort=-v:refname https://github.com/vllm-project/vllm-ascend "v*" | cut -f2 | sed 's|refs/tags/||')
83+
LATEST_TAG=$(echo "$TAGS" | head -n1)
84+
if [[ -z "$LATEST_TAG" ]]; then
85+
RESOLVED_VERSION="main"
86+
else
87+
RESOLVED_VERSION="$LATEST_TAG"
88+
fi
89+
else
90+
RESOLVED_VERSION="$VERSION_INPUT"
91+
fi
92+
echo "GHA_VLLM_ASCEND_VERSION=$RESOLVED_VERSION" >> $GITHUB_ENV
93+
94+
- name: Checkout vllm-project/vllm-ascend repo
95+
uses: actions/checkout@v4
96+
with:
97+
repository: vllm-project/vllm-ascend
98+
path: ./vllm-ascend
99+
ref: ${{ env.GHA_VLLM_ASCEND_VERSION }}
100+
101+
- name: Install vllm-project/vllm-ascend
102+
working-directory: ./vllm-ascend
103+
env:
104+
PIP_EXTRA_INDEX_URL: https://mirrors.huaweicloud.com/ascend/repos/pypi
105+
run: |
106+
pip install -r requirements-dev.txt
107+
pip install -v -e .
108+
109+
- name: Get vLLM commit hash and URL
110+
working-directory: ./vllm-empty
111+
run: |
112+
VLLM_COMMIT=$(git rev-parse --short=7 HEAD)
113+
echo "VLLM_COMMIT=$VLLM_COMMIT" >> $GITHUB_ENV
114+
115+
- name: Get vLLM-Ascend commit hash and URL
116+
working-directory: ./vllm-ascend
117+
run: |
118+
VLLM_ASCEND_COMMIT=$(git rev-parse --short=7 HEAD)
119+
echo "VLLM_ASCEND_COMMIT=$VLLM_ASCEND_COMMIT" >> $GITHUB_ENV
120+
121+
- name: Collect version info
122+
run: |
123+
for dir in /usr/local/Ascend/ascend-toolkit/*; do
124+
dname=$(basename "$dir")
125+
if [ "$dname" != "latest" ]; then
126+
TOOLKIT_DIR="$dname"
127+
break
128+
fi
129+
done
130+
INFO_FILE="/usr/local/Ascend/ascend-toolkit/${TOOLKIT_DIR}/$(uname -i)-linux/ascend_toolkit_install.info"
131+
GHA_CANN_VERSION=$(grep "version=" "$INFO_FILE" \
132+
| head -n1 \
133+
| cut -d'=' -f2 \
134+
| tr -d '"')
135+
{
136+
echo "GHA_CANN_VERSION=$GHA_CANN_VERSION"
137+
pip show torch | grep "Version:" | awk '{print "GHA_TORCH_VERSION="$2}'
138+
pip show torch_npu | grep "Version:" | awk '{print "GHA_TORCH_NPU_VERSION="$2}'
139+
pip show vllm | grep "Version:" | awk '{print "GHA_VLLM_VERSION="$2}' | sed 's/+.*//'
140+
} >> "$GITHUB_ENV"
141+
142+
- name: Run accuracy test
143+
id: report
144+
env:
145+
VLLM_WORKER_MULTIPROC_METHOD: spawn
146+
VLLM_USE_MODELSCOPE: True
147+
VLLM_VERSION: ${{ env.GHA_VLLM_VERSION }}
148+
VLLM_COMMIT: ${{ env.VLLM_COMMIT }}
149+
VLLM_ASCEND_VERSION: ${{ env.GHA_VLLM_ASCEND_VERSION || github.ref }}
150+
VLLM_ASCEND_COMMIT: ${{ env.VLLM_ASCEND_COMMIT }}
151+
CANN_VERSION: ${{ env.GHA_CANN_VERSION }}
152+
TORCH_VERSION: ${{ env.GHA_TORCH_VERSION }}
153+
TORCH_NPU_VERSION: ${{ env.GHA_TORCH_NPU_VERSION }}
154+
run: |
155+
model_base_name=$(basename ${{ inputs.model_name }})
156+
markdown_name="${model_base_name}"
157+
echo "markdown_name=$markdown_name" >> $GITHUB_OUTPUT
158+
mkdir -p ./benchmarks/accuracy
159+
pytest -sv ./tests/e2e/models/test_lm_eval_correctness.py \
160+
--config ./tests/e2e/models/configs/${{ inputs.model_name }}.yaml
161+
162+
- name: Generate step summary
163+
if: ${{ always() }}
164+
run: |
165+
cat ./benchmarks/accuracy/${{ steps.report.outputs.markdown_name }}.md >> $GITHUB_STEP_SUMMARY
166+
167+
- name: Upload Report
168+
if: ${{ inputs.upload == true }}
169+
uses: actions/upload-artifact@v4
170+
with:
171+
name: "report-${{ env.GHA_VLLM_ASCEND_VERSION }}-${{ steps.report.outputs.markdown_name }}"
172+
path: ./benchmarks/accuracy/${{ steps.report.outputs.markdown_name }}.md
173+
if-no-files-found: warn
174+
retention-days: 90
175+
overwrite: true

‎.github/workflows/_e2e_nightly.yaml‎

Lines changed: 115 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,115 @@
1+
#
2+
# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
3+
#
4+
# Licensed under the Apache License, Version 2.0 (the "License");
5+
# you may not use this file except in compliance with the License.
6+
# You may obtain a copy of the License at
7+
#
8+
# http://www.apache.org/licenses/LICENSE-2.0
9+
#
10+
# Unless required by applicable law or agreed to in writing, software
11+
# distributed under the License is distributed on an "AS IS" BASIS,
12+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13+
# See the License for the specific language governing permissions and
14+
# limitations under the License.
15+
# This file is a part of the vllm-ascend project.
16+
#
17+
18+
name: 'e2e nightly test'
19+
20+
on:
21+
workflow_call:
22+
inputs:
23+
vllm:
24+
required: true
25+
type: string
26+
runner:
27+
required: true
28+
type: string
29+
image:
30+
required: false
31+
type: string
32+
default: "swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/cann:8.2.rc1-910b-ubuntu22.04-py3.11"
33+
tests:
34+
required: true
35+
type: string
36+
37+
# Bash shells do not use ~/.profile or ~/.bashrc so these shells need to be explicitly
38+
# declared as "shell: bash -el {0}" on steps that need to be properly activated.
39+
# It's used to activate ascend-toolkit environment variables.
40+
defaults:
41+
run:
42+
shell: bash -el {0}
43+
44+
# only cancel in-progress runs of the same workflow
45+
# and ignore the lint / 1 card / 4 cards test type
46+
concurrency:
47+
group: ${{ github.workflow }}-${{ github.ref }}
48+
cancel-in-progress: true
49+
50+
jobs:
51+
e2e-nightly:
52+
name: e2e-nightly
53+
runs-on: ${{ inputs.runner }}
54+
container:
55+
image: ${{ inputs.image }}
56+
env:
57+
VLLM_USE_MODELSCOPE: True
58+
steps:
59+
- name: Check npu and CANN info
60+
run: |
61+
npu-smi info
62+
cat /usr/local/Ascend/ascend-toolkit/latest/"$(uname -i)"-linux/ascend_toolkit_install.info
63+
64+
- name: Config mirrors
65+
run: |
66+
sed -i 's|ports.ubuntu.com|mirrors.tuna.tsinghua.edu.cn|g' /etc/apt/sources.list
67+
pip config set global.index-url https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple
68+
apt-get update -y
69+
apt install git -y
70+
git config --global url."https://gh-proxy.test.osinfra.cn/https://github.com/".insteadOf https://github.com/
71+
72+
- name: Checkout vllm-project/vllm-ascend repo
73+
uses: actions/checkout@v4
74+
75+
- name: Install system dependencies
76+
run: |
77+
apt-get -y install `cat packages.txt`
78+
apt-get -y install gcc g++ cmake libnuma-dev
79+
80+
- name: Checkout vllm-project/vllm repo
81+
uses: actions/checkout@v4
82+
with:
83+
repository: vllm-project/vllm
84+
ref: ${{ inputs.vllm }}
85+
path: ./vllm-empty
86+
87+
- name: Install vllm-project/vllm from source
88+
working-directory: ./vllm-empty
89+
run: |
90+
VLLM_TARGET_DEVICE=empty pip install -e .
91+
92+
- name: Install vllm-project/vllm-ascend
93+
env:
94+
PIP_EXTRA_INDEX_URL: https://mirrors.huaweicloud.com/ascend/repos/pypi
95+
run: |
96+
pip install -r requirements-dev.txt
97+
pip install -v -e .
98+
99+
- name: Checkout aisbench repo and Install aisbench
100+
run: |
101+
git clone https://gitee.com/aisbench/benchmark.git
102+
cd benchmark
103+
git checkout v3.0-20250930-master
104+
pip3 install -e ./
105+
pip3 install -r requirements/api.txt
106+
pip3 install -r requirements/extra.txt
107+
108+
- name: Run vllm-project/vllm-ascend test
109+
env:
110+
VLLM_WORKER_MULTIPROC_METHOD: spawn
111+
VLLM_USE_MODELSCOPE: True
112+
VLLM_CI_RUNNER: ${{ inputs.runner }}
113+
run: |
114+
# TODO: enable more tests
115+
pytest -sv ${{ inputs.tests }}

0 commit comments

Comments
 (0)