minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[NPU] Improve the execution efficiency and maintainability of pr‑test‑npu (#33724)

Browse Source 
Co-authored-by: Even Zhou <even.y.zhou@outlook.com>
Co-authored-by: sglang-npu-bot <sglangnpu@163.com>
This commit is contained in:
Cherry_ming
2026-08-08 16:42:44 +08:00
committed by GitHub
co-authored by Even Zhou sglang-npu-bot
parent e732c0a9dc
commit dd5d82bead
60 changed files with 978 additions and 565 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/workflows/_npu-pr-test-stage.yml
+213
View File
@@ -0,0 +1,213 @@
name: PR Test Stage for NPU
# Reusable workflow for one CUDA test stage. Caller pr-test-npu.yml forwards
on:
workflow_call:
inputs:
self_name:
description: 'Caller job key; used for partitions[suite] lookup.'
type: string
required: true
runner_config:
description: 'Key in scripts/ci/runner_configs.yml. Resolves install script, artifact version, install timeout, runs-on label, and rdma_devices.'
type: string
required: true
image:
description: 'The container image for NPU test environment.'
type: string
required: true
npu_device_type:
description: 'NPU device type, e.g. 910b, a3.'
type: string
default: 'a3'
run_timeout_minutes:
description: 'timeout-minutes for the Run test step. Required so compute_partitions.py can read it from pr-test.yml without a duplicated default constant.'
type: string
required: true
timeout_per_file:
description: 'run_suite.py --timeout-per-file value (empty = unset).'
type: string
default: ''
partitions:
description: 'partitions config, e.g. {"size":1,"arr":[0]}'
type: string
default: '{"size":1,"arr":[0]}'
ref:
description: 'Git ref (branch, tag, or SHA) to test. If not provided, uses the default branch.'
type: string
default: ''
skip_pr_test_health_check:
description: 'Git ref (branch, tag, or SHA) to test. If not provided, uses the default branch.'
type: string
default: 'false'
github-token:
description: 'GitHub token for API calls'
type: string
default: ${{ github.token }}
env:
SKIP_PR_TEST_HEALTH_CHECK: ${{ inputs.skip_pr_test_health_check }}
SGLANG_USE_MODELSCOPE: true
SGLANG_IS_IN_CI: true
HF_ENDPOINT: https://hf-mirror.com
TORCH_EXTENSIONS_DIR: /tmp/torch_extensions
PYTORCH_NPU_ALLOC_CONF: "expandable_segments:True"
STREAMS_PER_DEVICE: 32
jobs:
run:
runs-on: ${{ inputs.runner_config }}
container:
image: ${{ inputs.image }}
timeout-minutes: 240
strategy:
fail-fast: false
matrix:
partition: ${{ fromJson(inputs.partitions).arr }}
steps:
- name: Checkout code
uses: actions/checkout@v4
with:
ref: ${{ inputs.ref || github.ref }}
- name: Mark repository safe
run: |
git config --system --add safe.directory ${GITHUB_WORKSPACE}
- name: Check PR test health
uses: actions/github-script@v8
env:
SKIP_PR_TEST_HEALTH_CHECK: ${{ env.SKIP_PR_TEST_HEALTH_CHECK }}
with:
github-token: ${{ inputs.github-token || github.token }}
script: |
core.notice(`[health-check] START — event=${context.eventName}, runId=${context.runId}`);
// Skip when explicitly requested via env var (e.g. release branch cut)
if (process.env.SKIP_PR_TEST_HEALTH_CHECK === 'true') {
core.notice('[health-check] SKIP: SKIP_PR_TEST_HEALTH_CHECK=true');
return;
}
// Skip for scheduled runs — they should collect all failures, not fast-fail
if (context.eventName === 'schedule') {
core.notice('[health-check] SKIP: scheduled run');
return;
}
// Check lint status from the separate Lint workflow (lint.yml).
// listJobsForWorkflowRun only sees jobs within the SAME run, so we use
// checks.listForRef which queries by commit SHA across ALL workflows.
const ref = context.payload.pull_request?.head?.sha || context.sha;
core.info(`[health-check] Checking lint for ref=${ref}`);
const { data } = await github.rest.checks.listForRef({
owner: context.repo.owner,
repo: context.repo.repo,
ref: ref,
check_name: 'lint',
});
const lintRun = data.check_runs.find(
cr => cr.app?.slug === 'github-actions'
);
core.info(`[health-check] Lint check: status=${lintRun?.status}, conclusion=${lintRun?.conclusion}`);
if (lintRun?.status === 'completed' && lintRun?.conclusion === 'failure') {
core.setFailed('Fast-fail: lint check failed');
return;
}
// Skip the jobs-failed check when the PR carries the bypass-fastfail label.
// Lint check above still runs.
let labels = [];
if (context.payload.pull_request?.labels) {
labels = context.payload.pull_request.labels.map(l => l.name);
} else {
const { data: prs } = await github.rest.repos.listPullRequestsAssociatedWithCommit({
owner: context.repo.owner,
repo: context.repo.repo,
commit_sha: ref,
});
if (prs.length > 0) {
labels = prs[0].labels.map(l => l.name);
}
}
core.info(`[health-check] PR labels: [${labels.join(', ')}]`);
if (labels.includes('bypass-fastfail')) {
core.notice('[health-check] SKIP jobs-failed check: bypass-fastfail label present');
return;
}
const jobs = await github.paginate(github.rest.actions.listJobsForWorkflowRun, {
owner: context.repo.owner,
repo: context.repo.repo,
run_id: context.runId,
per_page: 100,
});
core.info(`[health-check] Total jobs in run: ${jobs.length}`);
const failedJobs = jobs.filter(j => j.status === 'completed' && j.conclusion === 'failure');
core.info(`[health-check] Failed jobs (before filtering): ${failedJobs.map(j => `${j.name}(${j.conclusion})`).join(', ') || 'none'}`);
// Find jobs that failed from a real error, not from fast-fail cascade
const rootCauseFailures = jobs.filter(j => {
if (j.status !== 'completed' || j.conclusion !== 'failure') return false;
// h20 runners are flaky (dirty GPU state from prior runs); their failures
// should not cascade fast-fail to other stages. j.name shape from
// listJobsForWorkflowRun: "<job-key>" + optional " / <reusable-job>"
// + optional " (<matrix>)". Split off the base job key before exact
// match so we cover both inline + reusable forms without confusing
// 'h20' with the 'h200' prefix.
const baseName = j.name.split(/[ /]/)[0];
if (baseName === 'base-c-test-8-gpu-h20') {
core.info(`[health-check] Filtered out h20 job: ${j.name}`);
return false;
}
// If the failing step is the health check, it's a cascade — skip it
const failedStep = (j.steps || []).find(s => s.conclusion === 'failure');
if (failedStep && (failedStep.name.includes('check-pr-test-health') || failedStep.name.includes('Check PR test health'))) {
core.info(`[health-check] Filtered out cascade failure: ${j.name} (failed step: ${failedStep.name})`);
return false;
}
return true;
});
core.info(`[health-check] Root cause failures (after filtering): ${rootCauseFailures.map(j => j.name).join(', ') || 'none'}`);
if (rootCauseFailures.length > 0) {
core.setFailed(`Fast-fail: skipping — root cause job(s): ${rootCauseFailures.map(j => j.name).join(', ')}`);
} else {
core.notice('[health-check] PASS: no root cause failures detected');
}
- name: Install dependencies
env:
TORCH_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu"
PYPI_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
UV_INDEX_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/"
RUSTUP_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local:8082"
run: |
# speed up by using infra cache services
CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local"
sed -Ei "s@(ports|archive).ubuntu.com@${CACHING_URL}:8081@g" /etc/apt/sources.list
pip config set global.index-url http://${CACHING_URL}/pypi/simple
pip config set global.trusted-host "${CACHING_URL}"
bash scripts/ci/npu/npu_ci_install_dependency.sh ${{ inputs.npu_device_type }}
# copy required file from our daily cache
cp ~/.cache/modelscope/hub/datasets/otavia/ShareGPT_Vicuna_unfiltered/ShareGPT_V3_unfiltered_cleaned_split.json /tmp
# copy gsm8k dataset
cp ~/.cache/modelscope/hub/datasets/tmp/test.jsonl /tmp
# install sglang_router
apt-get install -y libssl-dev
pip install sglang_router
- name: Run test
timeout-minutes: ${{ fromJson(inputs.run_timeout_minutes) }}
env:
CONTINUE_ON_ERROR_FLAG: ${{ inputs.continue_on_error == 'true' && '--continue-on-error' || '' }}
run: |
cd test
python3 run_suite.py --hw npu --suite ${{ inputs.self_name }} \
--auto-partition-id ${{ matrix.partition }} \
--auto-partition-size ${{ fromJson(inputs.partitions).size }} \
${{ inputs.timeout_per_file && format('--timeout-per-file {0}', inputs.timeout_per_file) || '' }} \
$CONTINUE_ON_ERROR_FLAG
.github/workflows/_npu-single-node-test-stage.yml
+295
View File
@@ -0,0 +1,295 @@
name: 'Single Node Template for E2E performance and accuracy tests'
on:
workflow_call:
inputs:
runner:
required: true
type: string
default: linux-aarch64-a3-16
test_type:
required: true
type: string
default: perf
description: perf or accuracy
test_suite:
required: true
type: string
default: ''
description: name of test suite to run via run_suite.py (mutually exclusive with test_case)
image:
required: true
type: string
description: image for pods
default: "swr.cn-southwest-2.myhuaweicloud.com/base_image/dockerhub/lmsysorg/sglang:main-cann9.0.0-a3"
install_sglang_deps:
required: false
type: boolean
default: false
description: install sglang dependencies (e.g. PyTorch, CANN packages) when using source installation
device_type_for_deps:
required: false
type: string
default: 'a3'
description: device type for dependency installation (a3 or 910b)
transformers_version:
required: false
type: string
default: ""
description: "The transformers version number for running sglang. Use default version in image if keep empty."
skip_pr_test_health_check:
description: 'Git ref (branch, tag, or SHA) to test. If not provided, uses the default branch.'
type: string
default: 'false'
github-token:
description: 'GitHub token for API calls'
type: string
default: ${{ github.token }}
env:
SKIP_PR_TEST_HEALTH_CHECK: ${{ inputs.skip_pr_test_health_check }}
concurrency:
group: ascend-nightly-e2e-singlenode-${{ github.workflow_ref }}-${{ github.ref }}-${{ inputs.test_suite }}
cancel-in-progress: true
jobs:
e2e:
name: ${{ inputs.test_suite }}
runs-on: ${{ inputs.runner }}
container:
image: ${{ inputs.image }}
steps:
- name: Checkout code
uses: actions/checkout@v4
- name: Check PR test health
uses: actions/github-script@v8
env:
SKIP_PR_TEST_HEALTH_CHECK: ${{ env.SKIP_PR_TEST_HEALTH_CHECK }}
with:
github-token: ${{ inputs.github-token || github.token }}
script: |
core.notice(`[health-check] START — event=${context.eventName}, runId=${context.runId}`);
// Skip when explicitly requested via env var (e.g. release branch cut)
if (process.env.SKIP_PR_TEST_HEALTH_CHECK === 'true') {
core.notice('[health-check] SKIP: SKIP_PR_TEST_HEALTH_CHECK=true');
return;
}
// Skip for scheduled runs — they should collect all failures, not fast-fail
if (context.eventName === 'schedule') {
core.notice('[health-check] SKIP: scheduled run');
return;
}
// Check lint status from the separate Lint workflow (lint.yml).
// listJobsForWorkflowRun only sees jobs within the SAME run, so we use
// checks.listForRef which queries by commit SHA across ALL workflows.
const ref = context.payload.pull_request?.head?.sha || context.sha;
core.info(`[health-check] Checking lint for ref=${ref}`);
const { data } = await github.rest.checks.listForRef({
owner: context.repo.owner,
repo: context.repo.repo,
ref: ref,
check_name: 'lint',
});
const lintRun = data.check_runs.find(
cr => cr.app?.slug === 'github-actions'
);
core.info(`[health-check] Lint check: status=${lintRun?.status}, conclusion=${lintRun?.conclusion}`);
if (lintRun?.status === 'completed' && lintRun?.conclusion === 'failure') {
core.setFailed('Fast-fail: lint check failed');
return;
}
// Skip the jobs-failed check when the PR carries the bypass-fastfail label.
// Lint check above still runs.
let labels = [];
if (context.payload.pull_request?.labels) {
labels = context.payload.pull_request.labels.map(l => l.name);
} else {
const { data: prs } = await github.rest.repos.listPullRequestsAssociatedWithCommit({
owner: context.repo.owner,
repo: context.repo.repo,
commit_sha: ref,
});
if (prs.length > 0) {
labels = prs[0].labels.map(l => l.name);
}
}
core.info(`[health-check] PR labels: [${labels.join(', ')}]`);
if (labels.includes('bypass-fastfail')) {
core.notice('[health-check] SKIP jobs-failed check: bypass-fastfail label present');
return;
}
const jobs = await github.paginate(github.rest.actions.listJobsForWorkflowRun, {
owner: context.repo.owner,
repo: context.repo.repo,
run_id: context.runId,
per_page: 100,
});
core.info(`[health-check] Total jobs in run: ${jobs.length}`);
const failedJobs = jobs.filter(j => j.status === 'completed' && j.conclusion === 'failure');
core.info(`[health-check] Failed jobs (before filtering): ${failedJobs.map(j => `${j.name}(${j.conclusion})`).join(', ') || 'none'}`);
// Find jobs that failed from a real error, not from fast-fail cascade
const rootCauseFailures = jobs.filter(j => {
if (j.status !== 'completed' || j.conclusion !== 'failure') return false;
// h20 runners are flaky (dirty GPU state from prior runs); their failures
// should not cascade fast-fail to other stages. j.name shape from
// listJobsForWorkflowRun: "<job-key>" + optional " / <reusable-job>"
// + optional " (<matrix>)". Split off the base job key before exact
// match so we cover both inline + reusable forms without confusing
// 'h20' with the 'h200' prefix.
const baseName = j.name.split(/[ /]/)[0];
if (baseName === 'base-c-test-8-gpu-h20') {
core.info(`[health-check] Filtered out h20 job: ${j.name}`);
return false;
}
// If the failing step is the health check, it's a cascade — skip it
const failedStep = (j.steps || []).find(s => s.conclusion === 'failure');
if (failedStep && (failedStep.name.includes('check-pr-test-health') || failedStep.name.includes('Check PR test health'))) {
core.info(`[health-check] Filtered out cascade failure: ${j.name} (failed step: ${failedStep.name})`);
return false;
}
return true;
});
core.info(`[health-check] Root cause failures (after filtering): ${rootCauseFailures.map(j => j.name).join(', ') || 'none'}`);
if (rootCauseFailures.length > 0) {
core.setFailed(`Fast-fail: skipping — root cause job(s): ${rootCauseFailures.map(j => j.name).join(', ')}`);
} else {
core.notice('[health-check] PASS: no root cause failures detected');
}
- name: Check npu info
run: |
npu-smi info
- name: Install sglang dependencies
if: ${{ inputs.install_sglang_deps == true }}
shell: bash
env:
TORCH_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu"
PYPI_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
UV_INDEX_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/"
RUSTUP_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local:8082"
run: |
CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local"
sed -Ei "s@(ports|archive).ubuntu.com@${CACHING_URL}:8081@g" /etc/apt/sources.list
pip config set global.index-url http://${CACHING_URL}/pypi/simple
pip config set global.trusted-host "${CACHING_URL}"
bash scripts/ci/npu/npu_ci_install_dependency.sh ${{ inputs.device_type_for_deps }}
cp ~/.cache/modelscope/hub/datasets/otavia/ShareGPT_Vicuna_unfiltered/ShareGPT_V3_unfiltered_cleaned_split.json /tmp
curl -o /tmp/test.jsonl -L https://gh-proxy.test.osinfra.cn/https://raw.githubusercontent.com/openai/grade-school-math/master/grade_school_math/data/test.jsonl
- name: Run test
timeout-minutes: 300
env:
SGLANG_USE_MODELSCOPE: true
HF_ENDPOINT: https://hf-mirror.com
SGLANG_IS_IN_CI: true
TRANSFORMERS_VERBOSITY: "error"
GDN_ATTN_BACKEND_TRITON: 1
SGLANG_TEST_METRICS_OUTPUT: /root/.cache/tests/output/metrics/metrics
shell: bash
run: |
sglang_source_path=$(pwd)
echo "Source code path: ${sglang_source_path}"
ln -sf ${sglang_source_path} /root/sglang
# Determine test mode: suite (run_suite.py --suite) or single case file.
test_suite="${{ inputs.test_suite }}"
echo "Test mode: suite (${test_suite})"
tc_name="${test_suite}"
export TRANSFORMERS_VERSION_FOR_SGLANG="${{ inputs.transformers_version }}"
PYTHON_FOR_SGLANG="python"
PIP_FOR_SGLANG="pip"
if [ -n "${TRANSFORMERS_VERSION_FOR_SGLANG}" ];then
echo "===== Install transformers for sglang - Begin ====="
TRANSFORMERS_PKG_PATH_SOURCE=/root/.cache/.cache/transformers/${TRANSFORMERS_VERSION_FOR_SGLANG}
if [ ! -d "${TRANSFORMERS_PKG_PATH_SOURCE}" ]; then
echo "The dependent transformers package does not exist: ${TRANSFORMERS_PKG_PATH_SOURCE}."
echo "Install transformers ${TRANSFORMERS_VERSION_FOR_SGLANG} online."
pip install transformers=="${TRANSFORMERS_VERSION_FOR_SGLANG}" -i https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple
else
echo "Install transformers ${TRANSFORMERS_VERSION_FOR_SGLANG} locally."
TRANSFORMERS_PKG_PATH_TARGET=/tmp/transformers/${TRANSFORMERS_VERSION_FOR_SGLANG}
mkdir -p "${TRANSFORMERS_PKG_PATH_TARGET}"
cp "${TRANSFORMERS_PKG_PATH_SOURCE}/*" "${TRANSFORMERS_PKG_PATH_TARGET}/"
pip install --no-index --find-links="${TRANSFORMERS_PKG_PATH_TARGET}" transformers=="${TRANSFORMERS_VERSION_FOR_SGLANG}"
fi
echo "===== Install transformers for sglang in virtual env - End ====="
fi
echo "Transformers version for sglang: $(${PIP_FOR_SGLANG} show transformers | grep Version | cut -d: -f2)"
echo "scaling_governor performance num: \
$(cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor | grep performance | wc -l)"
echo "swappiness: $(cat /proc/sys/vm/swappiness)"
echo "numa_balancing: $(cat /proc/sys/kernel/numa_balancing)"
echo "sched_migration_cost_ns: $(cat /proc/sys/kernel/sched_migration_cost_ns)"
export SGLANG_TEST_MAX_RETRY=0
export SGLANG_SET_CPU_AFFINITY=1
echo "SGLANG_SET_CPU_AFFINITY: $SGLANG_SET_CPU_AFFINITY"
# Copy the checked‑out test common utility code to the sglang installation directory.
sglang_pkg_path=/sgl-workspace/sglang/python
ascend_test_util_path=${sglang_pkg_path}/sglang/test/ascend
mkdir -p ${ascend_test_util_path}
mv ${ascend_test_util_path} ${ascend_test_util_path}_bak
cp -r ${sglang_source_path}/python/sglang/test/ascend ${ascend_test_util_path}
source /usr/local/Ascend/cann/set_env.sh || true
source /usr/local/Ascend/nnal/atb/set_env.sh || true
source /usr/local/Ascend/ascend-toolkit/latest/opp/vendors/customize/bin/set_env.bash || true
source /usr/local/Ascend/ascend-toolkit/latest/opp/vendors/custom_transformer/bin/set_env.bash || true
# Set environment of cann
log_path="/root/.cache/tests/logs/log/${current_date}/${tc_name}/${HOSTNAME}"
rm -rf ${log_path}
mkdir -p ${log_path}
echo "Log path: ${log_path}"
echo "Running test: ${tc_name}"
test_exit_code=0
cd test
${PYTHON_FOR_SGLANG} -u run_suite.py --hw npu --suite ${test_suite} --timeout-per-file 3600 2>&1 | tee /tmp/test_output.log || test_exit_code=$?
echo "Finished test: ${tc_name}"
if [ "${test_exit_code}" = "0" ]; then
test_status="pass"
status_icon="✅"
else
test_status="fail"
status_icon="❌"
fi
echo "test_status=${test_status}" >> $GITHUB_ENV
echo "tc_name=${tc_name}" >> $GITHUB_ENV
export test_status tc_name
echo "## ${tc_name} ${status_icon} ${test_status}" >> $GITHUB_STEP_SUMMARY
echo "" >> $GITHUB_STEP_SUMMARY
metric_count=$(grep -c '\[METRIC\]' /tmp/test_output.log 2>/dev/null || echo 0)
if [ "${metric_count}" -gt 0 ]; then
echo "| Metric | Value | Pass |" >> $GITHUB_STEP_SUMMARY
echo "|--------|-------|------|" >> $GITHUB_STEP_SUMMARY
grep '\[METRIC\]' /tmp/test_output.log | while IFS= read -r line; do
metric_name=$(echo "$line" | sed -E 's/.*\[METRIC\] ([^=]+)=.*/\1/')
metric_value=$(echo "$line" | sed -E 's/.*\[METRIC\] [^=]+=([^ ]+).*/\1/')
echo "| ${metric_name} | ${metric_value} | ${status_icon} |" >> $GITHUB_STEP_SUMMARY
done
else
echo "No metrics collected (test may have failed before producing results)." >> $GITHUB_STEP_SUMMARY
fi
echo "" >> $GITHUB_STEP_SUMMARY
exit ${test_exit_code}
.github/workflows/pr-test-npu.yml
+171 -404
View File
@@ -90,303 +90,82 @@ jobs:
echo "CANN_image_a3=swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/cann:9.0.0-a3-ubuntu22.04-py3.11" >> $GITHUB_OUTPUT
echo "CANN_image_910b=swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/cann:9.0.0-910b-ubuntu22.04-py3.11" >> $GITHUB_OUTPUT
stage-a-unit-test-npu:
base-a-test-1-npu-a2:
needs: [check-changes, pr-gate, set-image-config]
if: needs.check-changes.outputs.main_package == 'true'
runs-on: linux-aarch64-a2-1
container:
if: ${{ !failure() && !cancelled() && needs.check-changes.outputs.main_package == 'true' }}
uses: ./.github/workflows/_npu-pr-test-stage.yml
with:
self_name: base-a-test-1-npu-a2
runner_config: linux-aarch64-a2-1
image: ${{ needs.set-image-config.outputs.CANN_image_910b }}
steps:
- name: Checkout code
uses: actions/checkout@v4
with:
ref: ${{ inputs.ref || github.ref }}
npu_device_type: 910b
run_timeout_minutes: '15'
secrets: inherit
- name: Mark repository safe
run: |
git config --system --add safe.directory ${GITHUB_WORKSPACE}
- name: Install dependencies
env:
TORCH_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu"
PYPI_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
UV_INDEX_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/"
RUSTUP_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local:8082"
run: |
# speed up by using infra cache services
CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local"
sed -Ei "s@(ports|archive).ubuntu.com@${CACHING_URL}:8081@g" /etc/apt/sources.list
pip config set global.index-url http://${CACHING_URL}/pypi/simple
pip config set global.trusted-host "${CACHING_URL}"
bash scripts/ci/npu/npu_ci_install_dependency.sh 910b
- name: Run test
timeout-minutes: 15
env:
SGLANG_IS_IN_CI: true
run: |
cd test
python3 run_suite.py --hw npu --suite stage-a-unit-test-npu
stage-b-test-1-npu-a3:
base-b-test-1-npu-a3:
needs: [check-changes, pr-gate, set-image-config]
if: needs.check-changes.outputs.main_package == 'true'
runs-on: linux-aarch64-a3-2-
container:
if: ${{ !failure() && !cancelled() && needs.check-changes.outputs.main_package == 'true' }}
uses: ./.github/workflows/_npu-pr-test-stage.yml
with:
self_name: base-b-test-1-npu-a3
runner_config: linux-aarch64-a3-2-
image: ${{ needs.set-image-config.outputs.CANN_image_a3 }}
steps:
- name: Checkout code
uses: actions/checkout@v4
with:
ref: ${{ inputs.ref || github.ref }}
run_timeout_minutes: '60'
timeout_per_file: '3600'
secrets: inherit
- name: Mark repository safe
run: |
git config --system --add safe.directory ${GITHUB_WORKSPACE}
- name: Install dependencies
env:
TORCH_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu"
PYPI_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
UV_INDEX_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/"
RUSTUP_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local:8082"
run: |
# speed up by using infra cache services
CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local"
sed -Ei "s@(ports|archive).ubuntu.com@${CACHING_URL}:8081@g" /etc/apt/sources.list
pip config set global.index-url http://${CACHING_URL}/pypi/simple
pip config set global.trusted-host "${CACHING_URL}"
bash scripts/ci/npu/npu_ci_install_dependency.sh a3
# copy required file from our daily cache
cp ~/.cache/modelscope/hub/datasets/otavia/ShareGPT_Vicuna_unfiltered/ShareGPT_V3_unfiltered_cleaned_split.json /tmp
# copy gsm8k dataset
cp ~/.cache/modelscope/hub/datasets/tmp/test.jsonl /tmp
- name: Run test
timeout-minutes: 60
env:
SGLANG_USE_MODELSCOPE: true
SGLANG_IS_IN_CI: true
HF_ENDPOINT: https://hf-mirror.com
TORCH_EXTENSIONS_DIR: /tmp/torch_extensions
PYTORCH_NPU_ALLOC_CONF: "expandable_segments:True"
STREAMS_PER_DEVICE: 32
run: |
cd test
python3 run_suite.py --hw npu --suite stage-b-test-1-npu-a3 --timeout-per-file 3600
stage-b-test-2-npu-a3:
base-b-test-2-npu-a3:
needs: [check-changes, pr-gate, set-image-config]
if: needs.check-changes.outputs.main_package == 'true'
runs-on: linux-aarch64-a3-2-
container:
if: ${{ !failure() && !cancelled() && needs.check-changes.outputs.main_package == 'true' }}
uses: ./.github/workflows/_npu-pr-test-stage.yml
with:
self_name: base-b-test-2-npu-a3
runner_config: linux-aarch64-a3-2-
image: ${{ needs.set-image-config.outputs.CANN_image_a3 }}
steps:
- name: Checkout code
uses: actions/checkout@v4
with:
ref: ${{ inputs.ref || github.ref }}
run_timeout_minutes: '60'
timeout_per_file: '3600'
secrets: inherit
- name: Mark repository safe
run: |
git config --system --add safe.directory ${GITHUB_WORKSPACE}
- name: Install dependencies
env:
TORCH_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu"
PYPI_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
UV_INDEX_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/"
RUSTUP_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local:8082"
run: |
# speed up by using infra cache services
CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local"
sed -Ei "s@(ports|archive).ubuntu.com@${CACHING_URL}:8081@g" /etc/apt/sources.list
pip config set global.index-url http://${CACHING_URL}/pypi/simple
pip config set global.trusted-host "${CACHING_URL}"
bash scripts/ci/npu/npu_ci_install_dependency.sh a3
# copy required file from our daily cache
cp ~/.cache/modelscope/hub/datasets/otavia/ShareGPT_Vicuna_unfiltered/ShareGPT_V3_unfiltered_cleaned_split.json /tmp
# copy gsm8k dataset
cp ~/.cache/modelscope/hub/datasets/tmp/test.jsonl /tmp
- name: Run test
timeout-minutes: 60
env:
SGLANG_USE_MODELSCOPE: true
SGLANG_IS_IN_CI: true
HF_ENDPOINT: https://hf-mirror.com
TORCH_EXTENSIONS_DIR: /tmp/torch_extensions
PYTORCH_NPU_ALLOC_CONF: "expandable_segments:True"
STREAMS_PER_DEVICE: 32
run: |
cd test
python3 run_suite.py --hw npu --suite stage-b-test-2-npu-a3 --timeout-per-file 3600
stage-b-test-4-npu-a3:
base-b-test-4-npu-a3:
needs: [check-changes, pr-gate, set-image-config]
if: needs.check-changes.outputs.main_package == 'true'
runs-on: linux-aarch64-a3-4-
strategy:
fail-fast: false
matrix:
part: [ 0, 1 ]
container:
if: ${{ !failure() && !cancelled() && needs.check-changes.outputs.main_package == 'true' }}
uses: ./.github/workflows/_npu-pr-test-stage.yml
with:
self_name: base-b-test-4-npu-a3
runner_config: linux-aarch64-a3-4-
image: ${{ needs.set-image-config.outputs.CANN_image_a3 }}
steps:
- name: Checkout code
uses: actions/checkout@v4
with:
ref: ${{ inputs.ref || github.ref }}
run_timeout_minutes: '120'
timeout_per_file: '3600'
partitions: '{"size":2,"arr":[0, 1]}'
secrets: inherit
- name: Mark repository safe
run: |
git config --system --add safe.directory ${GITHUB_WORKSPACE}
- name: Install dependencies
env:
TORCH_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu"
PYPI_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
UV_INDEX_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/"
RUSTUP_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local:8082"
run: |
# speed up by using infra cache services
CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local"
sed -Ei "s@(ports|archive).ubuntu.com@${CACHING_URL}:8081@g" /etc/apt/sources.list
pip config set global.index-url http://${CACHING_URL}/pypi/simple
pip config set global.trusted-host "${CACHING_URL}"
bash scripts/ci/npu/npu_ci_install_dependency.sh a3
# copy required file from our daily cache
cp ~/.cache/modelscope/hub/datasets/otavia/ShareGPT_Vicuna_unfiltered/ShareGPT_V3_unfiltered_cleaned_split.json /tmp
# copy gsm8k dataset
cp ~/.cache/modelscope/hub/datasets/tmp/test.jsonl /tmp
- name: Run test
timeout-minutes: 120
env:
SGLANG_USE_MODELSCOPE: true
SGLANG_IS_IN_CI: true
HF_ENDPOINT: https://hf-mirror.com
TORCH_EXTENSIONS_DIR: /tmp/torch_extensions
PYTORCH_NPU_ALLOC_CONF: "expandable_segments:True"
STREAMS_PER_DEVICE: 32
run: |
pip install sglang_router
cd test
python3 run_suite.py --hw npu --suite stage-b-test-4-npu-a3 --auto-partition-id ${{ matrix.part }} --auto-partition-size 2 --timeout-per-file 3600
stage-b-test-8-npu-a3:
base-b-test-8-npu-a3:
needs: [check-changes, pr-gate, set-image-config]
if: needs.check-changes.outputs.main_package == 'true'
runs-on: linux-aarch64-a3-8-
container:
if: ${{ !failure() && !cancelled() && needs.check-changes.outputs.main_package == 'true' }}
uses: ./.github/workflows/_npu-pr-test-stage.yml
with:
self_name: base-b-test-8-npu-a3
runner_config: linux-aarch64-a3-8-
image: ${{ needs.set-image-config.outputs.CANN_image_a3 }}
steps:
- name: Checkout code
uses: actions/checkout@v4
with:
ref: ${{ inputs.ref || github.ref }}
run_timeout_minutes: '60'
timeout_per_file: '3600'
secrets: inherit
- name: Mark repository safe
run: |
git config --system --add safe.directory ${GITHUB_WORKSPACE}
- name: Install dependencies
env:
TORCH_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu"
PYPI_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
UV_INDEX_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/"
RUSTUP_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local:8082"
run: |
# speed up by using infra cache services
CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local"
sed -Ei "s@(ports|archive).ubuntu.com@${CACHING_URL}:8081@g" /etc/apt/sources.list
pip config set global.index-url http://${CACHING_URL}/pypi/simple
pip config set global.trusted-host "${CACHING_URL}"
bash scripts/ci/npu/npu_ci_install_dependency.sh a3
# copy required file from our daily cache
cp ~/.cache/modelscope/hub/datasets/otavia/ShareGPT_Vicuna_unfiltered/ShareGPT_V3_unfiltered_cleaned_split.json /tmp
# copy gsm8k dataset
cp ~/.cache/modelscope/hub/datasets/tmp/test.jsonl /tmp
- name: Run test
timeout-minutes: 60
env:
SGLANG_USE_MODELSCOPE: true
SGLANG_IS_IN_CI: true
HF_ENDPOINT: https://hf-mirror.com
TORCH_EXTENSIONS_DIR: /tmp/torch_extensions
PYTORCH_NPU_ALLOC_CONF: "expandable_segments:True"
STREAMS_PER_DEVICE: 32
run: |
cd test
python3 run_suite.py --hw npu --suite stage-b-test-8-npu-a3 --timeout-per-file 3600
stage-b-test-16-npu-a3:
base-b-test-16-npu-a3:
needs: [check-changes, pr-gate, set-image-config]
if: needs.check-changes.outputs.main_package == 'true'
runs-on: linux-aarch64-a3-16-
container:
if: ${{ !failure() && !cancelled() && needs.check-changes.outputs.main_package == 'true' }}
uses: ./.github/workflows/_npu-pr-test-stage.yml
with:
self_name: base-b-test-16-npu-a3
runner_config: linux-aarch64-a3-16-
image: ${{ needs.set-image-config.outputs.CANN_image_a3 }}
steps:
- name: Checkout code
uses: actions/checkout@v4
with:
ref: ${{ inputs.ref || github.ref }}
- name: Mark repository safe
run: |
git config --system --add safe.directory ${GITHUB_WORKSPACE}
- name: Install dependencies
env:
TORCH_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu"
PYPI_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
UV_INDEX_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/"
RUSTUP_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local:8082"
run: |
# speed up by using infra cache services
CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local"
sed -Ei "s@(ports|archive).ubuntu.com@${CACHING_URL}:8081@g" /etc/apt/sources.list
pip config set global.index-url http://${CACHING_URL}/pypi/simple
pip config set global.trusted-host "${CACHING_URL}"
bash scripts/ci/npu/npu_ci_install_dependency.sh a3
# copy required file from our daily cache
cp ~/.cache/modelscope/hub/datasets/otavia/ShareGPT_Vicuna_unfiltered/ShareGPT_V3_unfiltered_cleaned_split.json /tmp
# copy gsm8k dataset
cp ~/.cache/modelscope/hub/datasets/tmp/test.jsonl /tmp
- name: Run test
timeout-minutes: 120
env:
SGLANG_USE_MODELSCOPE: true
SGLANG_IS_IN_CI: true
HF_ENDPOINT: https://hf-mirror.com
TORCH_EXTENSIONS_DIR: /tmp/torch_extensions
PYTORCH_NPU_ALLOC_CONF: "expandable_segments:True"
STREAMS_PER_DEVICE: 32
run: |
apt-get install -y libssl-dev
pip install sglang_router
cd test
python3 run_suite.py --hw npu --suite stage-b-test-16-npu-a3 --timeout-per-file 3600
run_timeout_minutes: '120'
timeout_per_file: '3600'
secrets: inherit
multimodal-gen-test-1-npu-a3:
needs: [check-changes, pr-gate, set-image-config]
if: needs.check-changes.outputs.multimodal_gen == 'true'
if: ${{ !failure() && !cancelled() && needs.check-changes.outputs.multimodal_gen == 'true' }}
runs-on: linux-aarch64-a3-800t-2
container:
image: ${{ needs.set-image-config.outputs.CANN_image_a3 }}
@@ -443,7 +222,7 @@ jobs:
multimodal-gen-test-2-npu-a3:
needs: [check-changes, pr-gate, set-image-config]
if: needs.check-changes.outputs.multimodal_gen == 'true'
if: ${{ !failure() && !cancelled() && needs.check-changes.outputs.multimodal_gen == 'true' }}
runs-on: linux-aarch64-a3-800t-2
container:
image: ${{ needs.set-image-config.outputs.CANN_image_a3 }}
@@ -498,146 +277,134 @@ jobs:
if-no-files-found: ignore
retention-days: 7
pr-single-node-tests:
name: single-node-poc
needs: [check-changes, pr-gate, set-image-config]
if: needs.check-changes.outputs.main_package == 'true'
strategy:
fail-fast: false
max-parallel: 6
matrix:
test_config:
# qwen3_6_27b performance tests
- name: qwen3_6_27b_w8a8_1p_in64k_out1k_50ms
runner: linux-aarch64-a3-800t-2
test_case: test/registered/npu/performance/qwen3_6_27b/test_npu_qwen3_6_27b_w8a8_1p_in64k_out1k_50ms.py
test_type: 'perf'
# - name: qwen3_6_27b_1p_in1024x1024_30_out1024_50ms
# runner: linux-aarch64-a3-800t-2
# test_case: test/registered/npu/performance/qwen3_6_27b/test_npu_qwen3_6_27b_1p_in1024x1024_30_out1024_50ms.py
# test_type: 'perf'
# # qwen3_8b performance tests
# - name: qwen3_8b_w8a8_1p_in3k5_out1k5_50ms
# runner: linux-aarch64-a3-800t-2
# test_case: test/registered/npu/performance/qwen3-8b/test_npu_qwen3_8b_w8a8_1p_in3k5_out1k5_50ms.py
# test_type: 'perf'
# # qwen3_30b_a3b performance tests
# - name: qwen3_30b_w8a8_1p_in3k5_out1k5_50ms
# runner: linux-aarch64-a3-800t-2
# test_case: test/registered/npu/performance/qwen3_30b_a3b/test_npu_qwen3_30b_w8a8_1p_in3k5_out1k5_50ms.py
# test_type: 'perf'
# # qwen3_6_35b_a3b performance tests
# - name: qwen3_6_35b_a3b_1p_in64k_out1k_prefix90_50ms
# runner: linux-aarch64-a3-800t-2
# test_case: test/registered/npu/performance/qwen3_6_35b_a3b/test_npu_qwen3_6_35b_a3b_1p_in64k_out1k_prefix90_50ms.py
# test_type: 'perf'
# # qwen3_vl_8b_thinking accuracy tests
# - name: qwen3_vl_8b_thinking_1p_mmmu
# runner: linux-aarch64-a3-2-
# test_case: test/registered/npu/accuracy/qwen3_vl_8b_thinking/test_npu_qwen3_vl_8b_thinking_1p_mmmu.py
# test_type: 'accuracy'
# # qwen3_32b performance tests
# - name: qwen3_32b_w8a8_2p_in3k5_out1k5_50ms
# runner: linux-aarch64-a3-800t-4
# test_case: test/registered/npu/performance/qwen3_32b/test_npu_qwen3_32b_w8a8_2p_in3k5_out1k5_50ms.py
# test_type: 'perf'
# # qwen3_next_80b_a3b performance tests
# - name: qwen3_next_80b_w8a8_2p_in6k_out1k5_bs16
# runner: linux-aarch64-a3-800t-4
# test_case: test/registered/npu/performance/qwen3_next_80b_a3b_instruct/test_npu_qwen3_next_80b_w8a8_2p_in6k_out1k5_bs16.py
# test_type: 'perf'
# # minimax_m2_5 performance tests
# - name: minimax_m2_5_w8a8_4p_in64k_out1k_prefix90_50ms
# runner: linux-aarch64-a3-800t-8
# test_case: test/registered/npu/performance/minimax_m2_5/test_npu_minimax_m2_5_w8a8_4p_in64k_out1k_prefix90_50ms.py
# test_type: 'perf'
# # deepseek_v4_flash performance tests
# - name: deepseek_v4_flash_w8a8_8p_in8k_out1k_50ms
# runner: linux-aarch64-a3-800t-16
# test_case: test/registered/npu/performance/deepseek_v4_flash/test_npu_deepseek_v4_flash_w8a8_8p_in8k_out1k_50ms.py
# test_type: 'perf'
# # kimi_k2_6 performance tests
# - name: kimi_k2_6_w4a8_8p_in3k5_out1k5_20ms
# runner: linux-aarch64-a3-800t-16
# test_case: test/registered/npu/performance/kimi_k2_6/test_npu_kimi_k2_6_w4a8_8p_in3k5_out1k5_20ms.py
# test_type: 'perf'
# # qwen3_235b performance tests
# - name: qwen3_235b_w8a8_8p_in3k5_out1k5_50ms
# runner: linux-aarch64-a3-800t-16
# test_case: test/registered/npu/performance/qwen3_235b_a22b/test_npu_qwen3_235b_w8a8_8p_in3k5_out1k5_50ms.py
# test_type: 'perf'
# # qwen3_5_397b performance tests
# - name: qwen3_5_397b_w4a8_8p_in3k5_out1k5_50ms
# runner: linux-aarch64-a3-800t-16
# test_case: test/registered/npu/performance/qwen3_5_397b/test_npu_qwen3_5_397b_w4a8_8p_in3k5_out1k5_50ms.py
# test_type: 'perf'
# NPU accuracy tests
# - name: glm4_7_flash_1p_gsm8k
# runner: linux-aarch64-a3-2-
# test_case: test/registered/npu/accuracy/glm4_7_flash/test_npu_glm4_7_flash_1p_gsm8k.py
# test_type: 'accuracy'
# - name: qwen3_vl_30b_a3b_bf16_2p_gsm8k
# runner: linux-aarch64-a3-4-
# test_case: test/registered/npu/accuracy/qwen3_vl_30b_a3b/test_npu_qwen3_vl_30b_a3b_bf16_2p_gsm8k.py
# test_type: 'accuracy'
- name: glm5_top64_pruned_bf16_8p_gsm8k
runner: linux-aarch64-a3-16-
test_case: test/registered/npu/accuracy/glm5_top64_pruned/test_npu_glm5_top64_pruned_bf16_8p_gsm8k.py
test_type: 'accuracy'
# - name: moonshotai_moonlight_16b_a3b_bf16_1p_gsm8k
# runner: linux-aarch64-a3-2-
# test_case: test/registered/npu/accuracy/moonshotai_moonlight_16b_a3b/test_npu_moonlight_16b_a3b_bf16_1p_gsm8k.py
# test_type: 'accuracy'
# - name: qwen3_5_9b_bf16_1p_gsm8k
# runner: linux-aarch64-a3-2-
# test_case: test/registered/npu/accuracy/qwen3_5_9b/test_npu_qwen3_5_9b_bf16_1p_gsm8k.py
# test_type: 'accuracy'
# - name: qwen3_vl_8b_bf16_2p_gsm8k
# runner: linux-aarch64-a3-4-
# test_case: test/registered/npu/accuracy/qwen3_vl_8b/test_npu_qwen3_vl_8b_bf16_2p_gsm8k.py
# test_type: 'accuracy'
uses: ./.github/workflows/nightly-test-npu-e2e-single-node.yml
base-c-test-acc-2-npu-a3:
name: base-c-test-acc-2-npu-a3
needs: [ check-changes, pr-gate, set-image-config ]
if: ${{ !failure() && !cancelled() && needs.check-changes.outputs.main_package == 'true' }}
uses: ./.github/workflows/_npu-single-node-test-stage.yml
with:
runner: ${{ matrix.test_config.runner }}
test_type: ${{ matrix.test_config.test_type }}
test_config_name: ${{ matrix.test_config.name }}
test_case: ${{ matrix.test_config.test_case }}
runner: linux-aarch64-a3-2-
test_type: 'accuracy'
test_suite: base-c-test-acc-2-npu-a3
image: ${{ needs.set-image-config.outputs.CANN_image_a3 }}
install_sglang_from_source: false
install_sglang_deps: true
device_type_for_deps: 'a3'
transformers_version: ''
base-c-test-acc-4-npu-a3:
name: base-c-test-acc-4-npu-a3
needs: [ check-changes, pr-gate, set-image-config ]
if: ${{ !failure() && !cancelled() && needs.check-changes.outputs.main_package == 'true' }}
uses: ./.github/workflows/_npu-single-node-test-stage.yml
with:
runner: linux-aarch64-a3-4-
test_type: 'accuracy'
test_suite: base-c-test-acc-4-npu-a3
image: ${{ needs.set-image-config.outputs.CANN_image_a3 }}
install_sglang_deps: true
device_type_for_deps: 'a3'
base-c-test-acc-8-npu-a3:
name: base-c-test-acc-8-npu-a3
needs: [ check-changes, pr-gate, set-image-config ]
if: ${{ !failure() && !cancelled() && needs.check-changes.outputs.main_package == 'true' }}
uses: ./.github/workflows/_npu-single-node-test-stage.yml
with:
runner: linux-aarch64-a3-8-
test_type: 'accuracy'
test_suite: base-c-test-acc-8-npu-a3
image: ${{ needs.set-image-config.outputs.CANN_image_a3 }}
install_sglang_deps: true
device_type_for_deps: 'a3'
base-c-test-acc-16-npu-a3:
name: base-c-test-acc-16-npu-a3
needs: [ check-changes, pr-gate, set-image-config ]
if: ${{ !failure() && !cancelled() && needs.check-changes.outputs.main_package == 'true' }}
uses: ./.github/workflows/_npu-single-node-test-stage.yml
with:
runner: linux-aarch64-a3-16-
test_type: 'accuracy'
test_suite: base-c-test-acc-16-npu-a3
image: ${{ needs.set-image-config.outputs.CANN_image_a3 }}
install_sglang_deps: true
device_type_for_deps: 'a3'
base-c-test-perf-2-npu-a3:
name: base-c-test-perf-2-npu-a3
needs: [ check-changes, pr-gate, set-image-config, base-c-test-acc-2-npu-a3 ]
if: ${{ !failure() && !cancelled() && needs.check-changes.outputs.main_package == 'true' }}
uses: ./.github/workflows/_npu-single-node-test-stage.yml
with:
runner: linux-aarch64-a3-800t-2
test_type: 'perf'
test_suite: base-c-test-perf-2-npu-a3
image: ${{ needs.set-image-config.outputs.CANN_image_a3 }}
install_sglang_deps: true
device_type_for_deps: 'a3'
base-c-test-perf-4-npu-a3:
name: base-c-test-perf-4-npu-a3
needs: [ check-changes, pr-gate, set-image-config, base-c-test-acc-4-npu-a3 ]
if: ${{ !failure() && !cancelled() && needs.check-changes.outputs.main_package == 'true' }}
uses: ./.github/workflows/_npu-single-node-test-stage.yml
with:
runner: linux-aarch64-a3-800t-4
test_type: 'perf'
test_suite: base-c-test-perf-4-npu-a3
image: ${{ needs.set-image-config.outputs.CANN_image_a3 }}
install_sglang_deps: true
device_type_for_deps: 'a3'
base-c-test-perf-8-npu-a3:
name: base-c-test-perf-8-npu-a3
needs: [ check-changes, pr-gate, set-image-config, base-c-test-acc-8-npu-a3 ]
if: ${{ !failure() && !cancelled() && needs.check-changes.outputs.main_package == 'true' }}
uses: ./.github/workflows/_npu-single-node-test-stage.yml
with:
runner: linux-aarch64-a3-800t-8
test_type: 'perf'
test_suite: base-c-test-perf-8-npu-a3
image: ${{ needs.set-image-config.outputs.CANN_image_a3 }}
install_sglang_deps: true
device_type_for_deps: 'a3'
base-c-test-perf-16-npu-a3:
name: base-c-test-perf-16-npu-a3
needs: [ check-changes, pr-gate, set-image-config, base-c-test-acc-16-npu-a3 ]
if: ${{ !failure() && !cancelled() && needs.check-changes.outputs.main_package == 'true' }}
uses: ./.github/workflows/_npu-single-node-test-stage.yml
with:
runner: linux-aarch64-a3-800t-16
test_type: 'perf'
test_suite: base-c-test-perf-16-npu-a3
image: ${{ needs.set-image-config.outputs.CANN_image_a3 }}
install_sglang_deps: true
device_type_for_deps: 'a3'
pr-test-npu-finish:
needs:
[
check-changes,
stage-a-unit-test-npu,
stage-b-test-1-npu-a3,
stage-b-test-2-npu-a3,
stage-b-test-4-npu-a3,
stage-b-test-8-npu-a3,
stage-b-test-16-npu-a3,
base-a-test-1-npu-a2,
base-b-test-1-npu-a3,
base-b-test-2-npu-a3,
base-b-test-4-npu-a3,
base-b-test-8-npu-a3,
base-b-test-16-npu-a3,
multimodal-gen-test-1-npu-a3,
multimodal-gen-test-2-npu-a3,
pr-single-node-tests,
base-c-test-acc-2-npu-a3,
base-c-test-acc-4-npu-a3,
base-c-test-acc-8-npu-a3,
base-c-test-acc-16-npu-a3,
base-c-test-perf-2-npu-a3,
base-c-test-perf-4-npu-a3,
base-c-test-perf-8-npu-a3,
base-c-test-perf-16-npu-a3,
]
if: always()
runs-on: ubuntu-latest