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

[XPU][CI] Fix empty nightly dashboard (#37800)

Browse Source
This commit is contained in:
ashwini rathi
2026-09-07 09:29:51 +08:00
committed by GitHub
parent e0a83a2215
commit 0afba909e7
4 changed files with 164 additions and 92 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/workflows/nightly-test-intel.yml
+59 -1
View File
@@ -59,6 +59,17 @@ jobs:
env:
GITHUB_WORKSPACE: ${{ github.workspace }}
- name: Overlay in-tree sglang test helpers onto image site-packages
run: |
# Bridges the up-to-24h window between merge and nightly image rebuild.
dest=$(docker exec ci_sglang_xpu /opt/venv/bin/python3 -c \
"import sglang, os; print(os.path.dirname(sglang.__file__))")
docker exec ci_sglang_xpu cp \
/sglang-checkout/python/sglang/test/ci/ci_utils.py \
"$dest/test/ci/ci_utils.py"
docker exec ci_sglang_xpu /opt/venv/bin/python3 -c \
"import sglang.test.ci.ci_utils; print('overlay ok')"
- name: HF login + install run_suite extras
timeout-minutes: 10
run: |
@@ -101,6 +112,7 @@ jobs:
cd /sglang-checkout/test &&
OLMOCR_BENCH_DIR=/sglang-checkout/olmOCR-bench/bench_data \
GITHUB_STEP_SUMMARY=/sglang-checkout/github_summary.md \
SGLANG_IS_IN_CI=true \
SGLANG_TEST_METRICS_FILE=/sglang-checkout/nightly-xpu-1-gpu-metrics.jsonl \
python3 run_suite.py --hw xpu --suite nightly-xpu-1-gpu --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }}
" || TEST_EXIT_CODE=$?
@@ -154,6 +166,17 @@ jobs:
env:
GITHUB_WORKSPACE: ${{ github.workspace }}
- name: Overlay in-tree sglang test helpers onto image site-packages
run: |
# Bridges the up-to-24h window between merge and nightly image rebuild.
dest=$(docker exec ci_sglang_xpu /opt/venv/bin/python3 -c \
"import sglang, os; print(os.path.dirname(sglang.__file__))")
docker exec ci_sglang_xpu cp \
/sglang-checkout/python/sglang/test/ci/ci_utils.py \
"$dest/test/ci/ci_utils.py"
docker exec ci_sglang_xpu /opt/venv/bin/python3 -c \
"import sglang.test.ci.ci_utils; print('overlay ok')"
- name: HF login + install run_suite extras
timeout-minutes: 10
run: |
@@ -185,6 +208,7 @@ jobs:
source /opt/venv/bin/activate &&
cd /sglang-checkout/test &&
GITHUB_STEP_SUMMARY=/sglang-checkout/github_summary.md \
SGLANG_IS_IN_CI=true \
SGLANG_TEST_METRICS_FILE=/sglang-checkout/nightly-xpu-2-gpu-metrics.jsonl \
python3 run_suite.py --hw xpu --suite nightly-xpu-2-gpu --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }}
" || TEST_EXIT_CODE=$?
@@ -238,6 +262,17 @@ jobs:
env:
GITHUB_WORKSPACE: ${{ github.workspace }}
- name: Overlay in-tree sglang test helpers onto image site-packages
run: |
# Bridges the up-to-24h window between merge and nightly image rebuild.
dest=$(docker exec ci_sglang_xpu /opt/venv/bin/python3 -c \
"import sglang, os; print(os.path.dirname(sglang.__file__))")
docker exec ci_sglang_xpu cp \
/sglang-checkout/python/sglang/test/ci/ci_utils.py \
"$dest/test/ci/ci_utils.py"
docker exec ci_sglang_xpu /opt/venv/bin/python3 -c \
"import sglang.test.ci.ci_utils; print('overlay ok')"
- name: HF login + install run_suite extras
timeout-minutes: 10
run: |
@@ -269,6 +304,7 @@ jobs:
source /opt/venv/bin/activate &&
cd /sglang-checkout/test &&
GITHUB_STEP_SUMMARY=/sglang-checkout/github_summary.md \
SGLANG_IS_IN_CI=true \
SGLANG_TEST_METRICS_FILE=/sglang-checkout/nightly-xpu-4-gpu-metrics.jsonl \
python3 run_suite.py --hw xpu --suite nightly-xpu-4-gpu --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }}
" || TEST_EXIT_CODE=$?
@@ -322,6 +358,17 @@ jobs:
env:
GITHUB_WORKSPACE: ${{ github.workspace }}
- name: Overlay in-tree sglang test helpers onto image site-packages
run: |
# Bridges the up-to-24h window between merge and nightly image rebuild.
dest=$(docker exec ci_sglang_xpu /opt/venv/bin/python3 -c \
"import sglang, os; print(os.path.dirname(sglang.__file__))")
docker exec ci_sglang_xpu cp \
/sglang-checkout/python/sglang/test/ci/ci_utils.py \
"$dest/test/ci/ci_utils.py"
docker exec ci_sglang_xpu /opt/venv/bin/python3 -c \
"import sglang.test.ci.ci_utils; print('overlay ok')"
- name: HF login + install run_suite extras
timeout-minutes: 10
run: |
@@ -348,16 +395,27 @@ jobs:
- name: Nightly Test (8-GPU XPU)
timeout-minutes: 480
run: |
touch github_summary.md
touch github_summary.md nightly-xpu-8-gpu-metrics.jsonl
docker exec ci_sglang_xpu bash -c "
source /opt/venv/bin/activate &&
cd /sglang-checkout/test &&
GITHUB_STEP_SUMMARY=/sglang-checkout/github_summary.md \
SGLANG_IS_IN_CI=true \
SGLANG_TEST_METRICS_FILE=/sglang-checkout/nightly-xpu-8-gpu-metrics.jsonl \
python3 run_suite.py --hw xpu --suite nightly-xpu-8-gpu --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }}
" || TEST_EXIT_CODE=$?
echo "$(<github_summary.md)" >> $GITHUB_STEP_SUMMARY || true
exit ${TEST_EXIT_CODE:-0}
- name: Upload per-model metrics jsonl
if: always()
uses: actions/upload-artifact@v4
with:
name: nightly-xpu-8-gpu-metrics
path: nightly-xpu-8-gpu-metrics.jsonl
if-no-files-found: warn
retention-days: 30
- name: Cleanup container
if: always()
run: |
.github/workflows/xpu-ci-job-monitor.yml
+2 -2
View File
@@ -288,10 +288,10 @@ jobs:
with open(snap_path) as f:
snapshot = json.load(f)
# Only look at nightly-test-intel jobs
# Snapshot writer emits `workflow` (filename), not `workflow_path`.
nightly = [
j for j in snapshot.get("jobs", [])
if j.get("workflow_path", "").endswith("nightly-test-intel.yml")
if (j.get("workflow") or "").endswith("nightly-test-intel.yml")
]
# status_summary style counts, per job name