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

[CI][XPU] Stabilize XPU CI: pin UMD/IGC, retry infra flakes, right-size EAGLE3 (#32438)

Browse Source 
Co-authored-by: Ma Mingfei <mingfei.ma@intel.com>
This commit is contained in:
ashwini rathi
2026-08-04 16:28:52 +08:00
committed by GitHub
co-authored by Ma Mingfei
parent 17d19081d9
commit 53804d609c
6 changed files with 137 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/workflows/pr-test-xpu.yml
+28 -2
View File
@@ -113,7 +113,7 @@ jobs:
- name: Run stage-a tests
timeout-minutes: 30
run: |
docker exec ci_sglang_xpu bash -c "source /opt/venv/bin/activate && cd /sglang-checkout/test && python3 run_suite.py --hw xpu --suite stage-a-test-1-gpu-xpu"
docker exec ci_sglang_xpu bash -c "source /opt/venv/bin/activate && cd /sglang-checkout/test && python3 run_suite.py --hw xpu --suite stage-a-test-1-gpu-xpu --enable-retry"
- name: Cleanup container
if: always()
@@ -136,6 +136,19 @@ jobs:
.pytest_cache || true
find . -type d -name "__pycache__" -prune -exec rm -rf {} + || true
find . -type f -name "*.pyc" -delete || true
# SIGTERM sglang and drain GPU context before `docker rm -f`;
# SIGKILL leaves the xe/GuC exec queue registered and triggers a
# GT reset (+ devcoredump) on B580.
if docker ps --format '{{.Names}}' | grep -qx ci_sglang_xpu; then
docker exec ci_sglang_xpu bash -c '
pkill -TERM -f "sglang|run_suite|python3.*test_" 2>/dev/null || true
for _ in $(seq 1 30); do
pgrep -f "sglang::|sglang.launch_server" >/dev/null || break
sleep 1
done
pkill -KILL -f "sglang|run_suite" 2>/dev/null || true
' || true
fi
docker rm -f ci_sglang_xpu || true
if [[ -n "${CI_SGLANG_XPU_IMAGE:-}" ]]; then
docker rmi -f "${CI_SGLANG_XPU_IMAGE}" || true
@@ -189,7 +202,7 @@ jobs:
- name: Run stage-b tests
timeout-minutes: 60
run: |
docker exec ci_sglang_xpu bash -c "source /opt/venv/bin/activate && cd /sglang-checkout/test && python3 run_suite.py --hw xpu --suite stage-b-test-1-gpu-xpu"
docker exec ci_sglang_xpu bash -c "source /opt/venv/bin/activate && cd /sglang-checkout/test && python3 run_suite.py --hw xpu --suite stage-b-test-1-gpu-xpu --enable-retry"
- name: Cleanup container
if: always()
@@ -212,6 +225,19 @@ jobs:
.pytest_cache || true
find . -type d -name "__pycache__" -prune -exec rm -rf {} + || true
find . -type f -name "*.pyc" -delete || true
# SIGTERM sglang and drain GPU context before `docker rm -f`;
# SIGKILL leaves the xe/GuC exec queue registered and triggers a
# GT reset (+ devcoredump) on B580.
if docker ps --format '{{.Names}}' | grep -qx ci_sglang_xpu; then
docker exec ci_sglang_xpu bash -c '
pkill -TERM -f "sglang|run_suite|python3.*test_" 2>/dev/null || true
for _ in $(seq 1 30); do
pgrep -f "sglang::|sglang.launch_server" >/dev/null || break
sleep 1
done
pkill -KILL -f "sglang|run_suite" 2>/dev/null || true
' || true
fi
docker rm -f ci_sglang_xpu || true
if [[ -n "${CI_SGLANG_XPU_IMAGE:-}" ]]; then
docker rmi -f "${CI_SGLANG_XPU_IMAGE}" || true