minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 13 Packages Projects Releases Wiki Activity

[XPU][NIGHTLY] Add 8 XPU nightly tests, enable 1-gpu suite (#30246)

Browse Source 
Co-authored-by: arathi-hlab <arathi-hlab@users.noreply.github.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
ashwini rathi
2026-07-20 16:15:50 +08:00
committed by GitHub
co-authored by arathi-hlab Claude Opus 4.8
parent 50c118704a
commit fafa302e41
13 changed files with 828 additions and 36 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/workflows/nightly-test-intel.yml
+81 -7
View File
@@ -35,13 +35,78 @@ concurrency:
jobs:
nightly-xpu-1-gpu:
# Placeholder: no models currently registered to nightly-xpu-1-gpu run on
# intel-bmg-nightly. Add a model test to this suite to re-enable.
if: false
if: github.repository == 'sgl-project/sglang'
runs-on: intel-bmg-nightly
env:
DOCKERHUB_INTEL_USERNAME: ${{ secrets.DOCKERHUB_INTEL_USERNAME }}
DOCKERHUB_INTEL_TOKEN: ${{ secrets.DOCKERHUB_INTEL_TOKEN }}
steps:
- name: Placeholder
run: echo "nightly-xpu-1-gpu has no validated models; skipping."
- name: Reset workspace ownership
run: |
docker run --rm -v "${{ github.workspace }}:/w" busybox:latest \
chown -R "$(id -u):$(id -g)" /w || true
- name: Checkout code
uses: actions/checkout@v4
with:
fetch-depth: 0
ref: ${{ inputs.ref || github.sha }}
- name: Start CI container (pull intel/sglang-dev:latest)
run: |
export HF_TOKEN="$(cat ~/huggingface_token.txt)"
bash scripts/ci/xpu/xpu_ci_start_container.sh
env:
GITHUB_WORKSPACE: ${{ github.workspace }}
- name: HF login + install run_suite extras
timeout-minutes: 10
run: |
docker exec ci_sglang_xpu /opt/venv/bin/python3 -m pip install --no-cache-dir tabulate
docker exec ci_sglang_xpu /opt/venv/bin/python3 -m pip install --no-cache-dir --no-deps xgrammar==0.1.33
docker exec ci_sglang_xpu /opt/venv/bin/python3 -m pip install --no-cache-dir "lm-eval==0.4.9"
docker exec ci_sglang_xpu /opt/venv/bin/python3 -m pip install --no-cache-dir pytest
# Diffusion extras (needed by test_xpu_flux2_dev, test_xpu_zimage_turbo).
# Pins mirror python/pyproject_xpu.toml [project.optional-dependencies.diffusion].
docker exec ci_sglang_xpu /opt/venv/bin/python3 -m pip install --no-cache-dir "sglang[diffusion]"
docker exec ci_sglang_xpu /bin/bash -c '/opt/venv/bin/hf auth login --token ${HF_TOKEN}'
- name: Download olmOCR-bench dataset (for test_deepseek_ocr_2_olmbench)
timeout-minutes: 30
run: |
# PyMuPDF renders the bench PDFs to images; bench_sglang.py errors
# every sample without it.
docker exec ci_sglang_xpu /opt/venv/bin/python3 -m pip install --no-cache-dir pymupdf
docker exec ci_sglang_xpu /bin/bash -c '
/opt/venv/bin/hf download --repo-type dataset allenai/olmOCR-bench \
--local-dir /sglang-checkout/olmOCR-bench'
- name: Nightly Test (1-GPU XPU)
timeout-minutes: 240
run: |
touch github_summary.md
docker exec ci_sglang_xpu bash -c "
source /opt/venv/bin/activate &&
cd /sglang-checkout/test &&
OLMOCR_BENCH_DIR=/sglang-checkout/olmOCR-bench/bench_data \
GITHUB_STEP_SUMMARY=/sglang-checkout/github_summary.md \
python3 run_suite.py --hw xpu --suite nightly-xpu-1-gpu --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }}
" || TEST_EXIT_CODE=$?
echo "$(<github_summary.md)" >> $GITHUB_STEP_SUMMARY || true
exit ${TEST_EXIT_CODE:-0}
- name: Cleanup container
if: always()
run: |
docker run --rm -v "${{ github.workspace }}:/w" busybox:latest \
chown -R "$(id -u):$(id -g)" /w || true
rm -rf test/result.jsonl test/results test/.pytest_cache .pytest_cache || true
find . -type d -name "__pycache__" -prune -exec rm -rf {} + || true
find . -type f -name "*.pyc" -delete || true
docker rm -f ci_sglang_xpu || true
if [[ -n "${CI_SGLANG_XPU_IMAGE:-}" ]]; then
docker rmi -f "${CI_SGLANG_XPU_IMAGE}" || true
fi
nightly-xpu-2-gpu:
if: github.repository == 'sgl-project/sglang'
@@ -74,10 +139,14 @@ jobs:
docker exec ci_sglang_xpu /opt/venv/bin/python3 -m pip install --no-cache-dir tabulate
docker exec ci_sglang_xpu /opt/venv/bin/python3 -m pip install --no-cache-dir --no-deps xgrammar==0.1.33
docker exec ci_sglang_xpu /opt/venv/bin/python3 -m pip install --no-cache-dir "lm-eval==0.4.9"
docker exec ci_sglang_xpu /opt/venv/bin/python3 -m pip install --no-cache-dir pytest
# Diffusion extras (needed by test_xpu_flux2_dev, test_xpu_zimage_turbo).
# Pins mirror python/pyproject_xpu.toml [project.optional-dependencies.diffusion].
docker exec ci_sglang_xpu /opt/venv/bin/python3 -m pip install --no-cache-dir "sglang[diffusion]"
docker exec ci_sglang_xpu /bin/bash -c '/opt/venv/bin/hf auth login --token ${HF_TOKEN}'
- name: Nightly Test (2-GPU XPU)
timeout-minutes: 60
timeout-minutes: 240
run: |
touch github_summary.md
docker exec ci_sglang_xpu bash -c "
@@ -133,10 +202,14 @@ jobs:
docker exec ci_sglang_xpu /opt/venv/bin/python3 -m pip install --no-cache-dir tabulate
docker exec ci_sglang_xpu /opt/venv/bin/python3 -m pip install --no-cache-dir --no-deps xgrammar==0.1.33
docker exec ci_sglang_xpu /opt/venv/bin/python3 -m pip install --no-cache-dir "lm-eval==0.4.9"
docker exec ci_sglang_xpu /opt/venv/bin/python3 -m pip install --no-cache-dir pytest
# Diffusion extras (needed by test_xpu_flux2_dev, test_xpu_zimage_turbo).
# Pins mirror python/pyproject_xpu.toml [project.optional-dependencies.diffusion].
docker exec ci_sglang_xpu /opt/venv/bin/python3 -m pip install --no-cache-dir "sglang[diffusion]"
docker exec ci_sglang_xpu /bin/bash -c '/opt/venv/bin/hf auth login --token ${HF_TOKEN}'
- name: Nightly Test (4-GPU XPU)
timeout-minutes: 120
timeout-minutes: 480
run: |
touch github_summary.md
docker exec ci_sglang_xpu bash -c "
@@ -164,6 +237,7 @@ jobs:
check-all-jobs:
if: always() && (github.repository == 'sgl-project/sglang' || github.event_name == 'workflow_dispatch')
needs:
- nightly-xpu-1-gpu
- nightly-xpu-2-gpu
- nightly-xpu-4-gpu
runs-on: ubuntu-latest