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

[Nightly] Nightly pre-build image (#7388)

Browse Source 
### What this PR does / why we need it?
This pull request refactor nightly image build and simplify the logic of
multi workflows.
1. Nightly image build become the prerequisite when the test are
triggered by `schedule` or `workflow_dispatch`
2. Simplify the pull request select case logic
3. Next step: Implement replaceable nightly tests. Specifically, if
nightly tests are manually triggered, they can accept any optional
docker image to meet the needs of different commits(Which means the
image is customizable).
### Does this PR introduce _any_ user-facing change?

### How was this patch tested?

- vLLM version: v0.17.0
- vLLM main:
4034c3d32e

---------

Signed-off-by: wangli <wangli858794774@gmail.com>
This commit is contained in:
Li Wang
2026-03-25 09:24:01 +08:00
committed by GitHub
parent 8977be1df3
commit 8e3f8bab57
8 changed files with 315 additions and 380 deletions
Download Patch File Download Diff File Expand all files Collapse all files

199
.github/workflows/schedule_nightly_test_a3.yaml vendored
View File

@@ -49,86 +49,30 @@ concurrency:
jobs:
parse-trigger:
name: Parse trigger and determine test scope
runs-on: linux-aarch64-a2b3-0
if: >-
github.event_name == 'schedule' ||
github.event_name == 'workflow_dispatch' ||
contains(github.event.pull_request.labels.*.name, 'nightly-test')
outputs:
should_run: ${{ steps.parse.outputs.should_run }}
test_filter: ${{ steps.parse.outputs.test_filter }}
is_pr_event: ${{ steps.parse.outputs.is_pr_event }}
steps:
- name: Parse trigger
id: parse
uses: actions/github-script@v7
with:
script: |
const eventName = context.eventName;
uses: ./.github/workflows/_parse_trigger.yaml
function parseNightlyComment(body) {
if (!body) return null;
const match = body.trim().match(/^\/nightly(?:\s+(.+))?$/m);
if (!match) return null;
const args = (match[1] || '').trim();
if (!args || args === 'all') return 'all';
// Wrap with commas for exact-name matching: ",name1,name2,"
return ',' + args.split(/\s+/).join(',') + ',';
}
// schedule / workflow_dispatch: run all tests with pre-built image
if (eventName === 'schedule' || eventName === 'workflow_dispatch') {
core.setOutput('should_run', 'true');
core.setOutput('test_filter', 'all');
core.setOutput('is_pr_event', 'false');
return;
}
// pull_request (labeled / synchronize)
if (eventName === 'pull_request') {
const labels = context.payload.pull_request.labels.map(l => l.name);
if (!labels.includes('nightly-test')) {
core.setOutput('should_run', 'false');
core.setOutput('test_filter', '');
core.setOutput('is_pr_event', 'true');
return;
}
// Search comments for latest /nightly command
const prNumber = context.payload.pull_request.number;
const comments = await github.paginate(github.rest.issues.listComments, {
owner: context.repo.owner,
repo: context.repo.repo,
issue_number: prNumber,
per_page: 100,
});
let testFilter = null;
for (let i = comments.length - 1; i >= 0; i--) {
const result = parseNightlyComment(comments[i].body);
if (result !== null) { testFilter = result; break; }
}
// No /nightly comment found: do not run any tests
if (testFilter === null) {
core.info('nightly-test label present but no /nightly comment found; skipping.');
core.setOutput('should_run', 'false');
core.setOutput('test_filter', '');
core.setOutput('is_pr_event', 'true');
return;
}
core.setOutput('should_run', 'true');
core.setOutput('test_filter', testFilter);
core.setOutput('is_pr_event', 'true');
return;
}
// Fallback
core.setOutput('should_run', 'false');
core.setOutput('test_filter', '');
core.setOutput('is_pr_event', 'false');
build-image:
name: Build nightly-a3 image
if: github.event_name == 'schedule' || github.event_name == 'workflow_dispatch'
uses: ./.github/workflows/_nightly_image_build.yaml
with:
target: a3
secrets:
HW_USERNAME: ${{ secrets.HW_USERNAME }}
HW_TOKEN: ${{ secrets.HW_TOKEN }}
GITEE_TOKEN: ${{ secrets.GITEE_TOKEN }}
multi-node-tests:
name: multi-node
needs: [parse-trigger]
if: always() && needs.parse-trigger.outputs.should_run == 'true'
needs: [parse-trigger, build-image]
if: >-
always() &&
needs.parse-trigger.outputs.run == 'true' &&
(needs.build-image.result == 'success' || needs.build-image.result == 'skipped')
strategy:
fail-fast: false
max-parallel: 2
@@ -140,15 +84,9 @@ jobs:
- name: multi-node-qwen3-dp
config_file_path: Qwen3-235B-A22B.yaml
size: 2
# - name: multi-node-dpsk-4node-pd
# config_file_path: DeepSeek-R1-W8A8.yaml
# size: 4
- name: multi-node-qwenw8a8-2node
config_file_path: Qwen3-235B-W8A8.yaml
size: 2
# - name: multi-node-deepseek-r1-w8a8-eplb
# config_file_path: DeepSeek-R1-W8A8-EPLB.yaml
# size: 4
- name: multi-node-qwenw8a8-2node-eplb
config_file_path: Qwen3-235B-W8A8-EPLB.yaml
size: 2
@@ -190,19 +128,12 @@ jobs:
replicas: 1
size: ${{ matrix.test_config.size }}
config_file_path: ${{ matrix.test_config.config_file_path }}
vllm_ascend_ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.ref_name }}
is_pr_test: >-
vllm_ascend_ref: ${{ needs.parse-trigger.outputs.ref }}
should_run: >-
${{
needs.parse-trigger.outputs.is_pr_event == 'true' && (
needs.parse-trigger.outputs.test_filter == 'all' ||
contains(needs.parse-trigger.outputs.test_filter, format(',{0},', matrix.test_config.name))
)
}}
is_run: >-
${{
needs.parse-trigger.outputs.should_run == 'true' && (
needs.parse-trigger.outputs.test_filter == 'all' ||
contains(needs.parse-trigger.outputs.test_filter, format(',{0},', matrix.test_config.name))
needs.parse-trigger.outputs.run == 'true' && (
needs.parse-trigger.outputs.filter == 'all' ||
contains(needs.parse-trigger.outputs.filter, format(',{0},', matrix.test_config.name))
)
}}
secrets:
@@ -210,44 +141,22 @@ jobs:
single-node-tests:
name: single-node
needs: [parse-trigger, multi-node-tests]
if: always() && needs.parse-trigger.outputs.should_run == 'true'
needs: [parse-trigger, build-image, multi-node-tests]
if: >-
always() &&
needs.parse-trigger.outputs.run == 'true' &&
(needs.build-image.result == 'success' || needs.build-image.result == 'skipped')
strategy:
fail-fast: false
matrix:
test_config:
# pytest-driven tests
- name: qwen3-30b-acc
os: linux-aarch64-a3-4
tests: tests/e2e/weekly/single_node/models/test_qwen3_30b_acc.py
uses: ./.github/workflows/_e2e_nightly_single_node.yaml
with:
runner: ${{ matrix.test_config.os }}
image: 'swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/vllm-ascend:nightly-a3'
tests: ${{ matrix.test_config.tests }}
name: ${{ matrix.test_config.name }}
is_pr_test: >-
${{
needs.parse-trigger.outputs.is_pr_event == 'true' && (
needs.parse-trigger.outputs.test_filter == 'all' ||
contains(needs.parse-trigger.outputs.test_filter, format(',{0},', matrix.test_config.name))
)
}}
is_run: >-
${{
needs.parse-trigger.outputs.should_run == 'true' && (
needs.parse-trigger.outputs.test_filter == 'all' ||
contains(needs.parse-trigger.outputs.test_filter, format(',{0},', matrix.test_config.name))
)
}}
single-node-yaml-tests:
name: single-node
if: always() && needs.parse-trigger.outputs.should_run == 'true'
needs: [parse-trigger, multi-node-tests]
strategy:
fail-fast: false
matrix:
test_config:
- name: custom-multi-ops
os: linux-aarch64-a3-16
tests: tests/e2e/nightly/single_node/ops/multicard_ops_a3/
# YAML-driven tests
- name: deepseek-r1-0528-w8a8
os: linux-aarch64-a3-16
@@ -316,51 +225,13 @@ jobs:
with:
runner: ${{ matrix.test_config.os }}
image: 'swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/vllm-ascend:nightly-a3'
tests: ${{ matrix.test_config.tests }}
config_file_path: ${{ matrix.test_config.config_file_path }}
name: ${{ matrix.test_config.name }}
is_pr_test: >-
should_run: >-
${{
needs.parse-trigger.outputs.is_pr_event == 'true' && (
needs.parse-trigger.outputs.test_filter == 'all' ||
contains(needs.parse-trigger.outputs.test_filter, format(',{0},', matrix.test_config.name))
)
}}
is_run: >-
${{
needs.parse-trigger.outputs.should_run == 'true' && (
needs.parse-trigger.outputs.test_filter == 'all' ||
contains(needs.parse-trigger.outputs.test_filter, format(',{0},', matrix.test_config.name))
)
}}
custom-ops-tests:
name: test ops
needs: [parse-trigger, multi-node-tests]
if: always() && needs.parse-trigger.outputs.should_run == 'true'
strategy:
fail-fast: false
matrix:
test_config:
- name: custom-multi-ops
os: linux-aarch64-a3-16
tests: tests/e2e/nightly/single_node/ops/multicard_ops_a3/
uses: ./.github/workflows/_e2e_nightly_single_node.yaml
with:
runner: ${{ matrix.test_config.os }}
image: 'swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/vllm-ascend:nightly-a3'
tests: ${{ matrix.test_config.tests }}
name: ${{ matrix.test_config.name }}
is_pr_test: >-
${{
needs.parse-trigger.outputs.is_pr_event == 'true' && (
needs.parse-trigger.outputs.test_filter == 'all' ||
contains(needs.parse-trigger.outputs.test_filter, format(',{0},', matrix.test_config.name))
)
}}
is_run: >-
${{
needs.parse-trigger.outputs.should_run == 'true' && (
needs.parse-trigger.outputs.test_filter == 'all' ||
contains(needs.parse-trigger.outputs.test_filter, format(',{0},', matrix.test_config.name))
needs.parse-trigger.outputs.run == 'true' && (
needs.parse-trigger.outputs.filter == 'all' ||
contains(needs.parse-trigger.outputs.filter, format(',{0},', matrix.test_config.name))
)
}}