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

[CI][NPU] use internal HTTP cache for Rust toolchain via RUSTUP_CACHE (#24801)

Browse Source
This commit is contained in:
Goalina
2026-05-13 09:58:07 +03:00
committed by GitHub
parent f2a90094c9
commit 10a005e4ea
2 changed files with 94 additions and 86 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/workflows/pr-test-npu.yml
+70 -77
View File
@@ -102,9 +102,8 @@ jobs:
TORCH_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu" 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" 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" UV_INDEX_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/" GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/"
RUSTUP_DIST_SERVER: "https://rsproxy.cn" RUSTUP_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local:8082"
RUSTUP_UPDATE_ROOT: "https://rsproxy.cn/rustup"
run: | run: |
# speed up by using infra cache services # speed up by using infra cache services
CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local" CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local"
@@ -156,9 +155,8 @@ jobs:
TORCH_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu" 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" 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" UV_INDEX_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/" GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/"
RUSTUP_DIST_SERVER: "https://rsproxy.cn" RUSTUP_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local:8082"
RUSTUP_UPDATE_ROOT: "https://rsproxy.cn/rustup"
run: | run: |
# speed up by using infra cache services # speed up by using infra cache services
CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local" CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local"
@@ -206,9 +204,8 @@ jobs:
TORCH_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu" 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" 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" UV_INDEX_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/" GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/"
RUSTUP_DIST_SERVER: "https://rsproxy.cn" RUSTUP_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local:8082"
RUSTUP_UPDATE_ROOT: "https://rsproxy.cn/rustup"
run: | run: |
# speed up by using infra cache services # speed up by using infra cache services
CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local" CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local"
@@ -257,9 +254,8 @@ jobs:
TORCH_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu" 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" 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" UV_INDEX_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/" GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/"
RUSTUP_DIST_SERVER: "https://rsproxy.cn" RUSTUP_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local:8082"
RUSTUP_UPDATE_ROOT: "https://rsproxy.cn/rustup"
run: | run: |
# speed up by using infra cache services # speed up by using infra cache services
CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local" CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local"
@@ -305,9 +301,8 @@ jobs:
TORCH_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu" 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" 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" UV_INDEX_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/" GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/"
RUSTUP_DIST_SERVER: "https://rsproxy.cn" RUSTUP_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local:8082"
RUSTUP_UPDATE_ROOT: "https://rsproxy.cn/rustup"
run: | run: |
# speed up by using infra cache services # speed up by using infra cache services
CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local" CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local"
@@ -327,25 +322,25 @@ jobs:
SGLANG_USE_MODELSCOPE: true SGLANG_USE_MODELSCOPE: true
SGLANG_IS_IN_CI: true SGLANG_IS_IN_CI: true
HF_ENDPOINT: https://hf-mirror.com HF_ENDPOINT: https://hf-mirror.com
TORCH_EXTENSIONS_DIR: /tmp/torch_extensions TORCH_EXTENSIONS_DIR: /tmp/torch_extensions
PYTORCH_NPU_ALLOC_CONF: "expandable_segments:True" PYTORCH_NPU_ALLOC_CONF: "expandable_segments:True"
STREAMS_PER_DEVICE: 32 STREAMS_PER_DEVICE: 32
SGLANG_DIFFUSION_ARTIFACT_DIR: ${{ github.workspace }}/diffusion-failures SGLANG_DIFFUSION_ARTIFACT_DIR: ${{ github.workspace }}/diffusion-failures
run: | run: |
export PATH="/usr/local/Ascend/8.3.RC1/compiler/bishengir/bin:${PATH}" export PATH="/usr/local/Ascend/8.3.RC1/compiler/bishengir/bin:${PATH}"
cd python cd python
python3 sglang/multimodal_gen/test/run_suite_npu.py --suite 1-npu python3 sglang/multimodal_gen/test/run_suite_npu.py --suite 1-npu
- name: Upload diffusion failure artifacts - name: Upload diffusion failure artifacts
if: always() if: always()
uses: actions/upload-artifact@v4 uses: actions/upload-artifact@v4
with: with:
name: diffusion-failures-npu-1-${{ github.run_attempt }} name: diffusion-failures-npu-1-${{ github.run_attempt }}
path: diffusion-failures/ path: diffusion-failures/
if-no-files-found: ignore if-no-files-found: ignore
retention-days: 7 retention-days: 7
multimodal-gen-test-2-npu-a3: multimodal-gen-test-2-npu-a3:
needs: [check-changes, pr-gate] needs: [check-changes, pr-gate]
if: needs.check-changes.outputs.multimodal_gen == 'true' if: needs.check-changes.outputs.multimodal_gen == 'true'
runs-on: linux-aarch64-a3-16 runs-on: linux-aarch64-a3-16
@@ -364,9 +359,8 @@ jobs:
TORCH_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu" 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" 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" UV_INDEX_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/" GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/"
RUSTUP_DIST_SERVER: "https://rsproxy.cn" RUSTUP_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local:8082"
RUSTUP_UPDATE_ROOT: "https://rsproxy.cn/rustup"
run: | run: |
# speed up by using infra cache services # speed up by using infra cache services
CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local" CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local"
@@ -386,25 +380,25 @@ jobs:
SGLANG_USE_MODELSCOPE: true SGLANG_USE_MODELSCOPE: true
SGLANG_IS_IN_CI: true SGLANG_IS_IN_CI: true
HF_ENDPOINT: https://hf-mirror.com HF_ENDPOINT: https://hf-mirror.com
TORCH_EXTENSIONS_DIR: /tmp/torch_extensions TORCH_EXTENSIONS_DIR: /tmp/torch_extensions
PYTORCH_NPU_ALLOC_CONF: "expandable_segments:True" PYTORCH_NPU_ALLOC_CONF: "expandable_segments:True"
STREAMS_PER_DEVICE: 32 STREAMS_PER_DEVICE: 32
SGLANG_DIFFUSION_ARTIFACT_DIR: ${{ github.workspace }}/diffusion-failures SGLANG_DIFFUSION_ARTIFACT_DIR: ${{ github.workspace }}/diffusion-failures
run: | run: |
export PATH="/usr/local/Ascend/8.3.RC1/compiler/bishengir/bin:${PATH}" export PATH="/usr/local/Ascend/8.3.RC1/compiler/bishengir/bin:${PATH}"
cd python cd python
python3 sglang/multimodal_gen/test/run_suite_npu.py --suite 2-npu python3 sglang/multimodal_gen/test/run_suite_npu.py --suite 2-npu
- name: Upload diffusion failure artifacts - name: Upload diffusion failure artifacts
if: always() if: always()
uses: actions/upload-artifact@v4 uses: actions/upload-artifact@v4
with: with:
name: diffusion-failures-npu-2-${{ github.run_attempt }} name: diffusion-failures-npu-2-${{ github.run_attempt }}
path: diffusion-failures/ path: diffusion-failures/
if-no-files-found: ignore if-no-files-found: ignore
retention-days: 7 retention-days: 7
multimodal-gen-test-8-npu-a3: multimodal-gen-test-8-npu-a3:
needs: [check-changes, pr-gate] needs: [check-changes, pr-gate]
if: needs.check-changes.outputs.multimodal_gen == 'true' if: needs.check-changes.outputs.multimodal_gen == 'true'
runs-on: linux-aarch64-a3-8 runs-on: linux-aarch64-a3-8
@@ -423,9 +417,8 @@ jobs:
TORCH_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu" 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" 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" UV_INDEX_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/" GITHUB_PROXY_URL: "https://gh-proxy.test.osinfra.cn/"
RUSTUP_DIST_SERVER: "https://rsproxy.cn" RUSTUP_CACHE_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local:8082"
RUSTUP_UPDATE_ROOT: "https://rsproxy.cn/rustup"
run: | run: |
# speed up by using infra cache services # speed up by using infra cache services
CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local" CACHING_URL="cache-service.nginx-pypi-cache.svc.cluster.local"
@@ -445,24 +438,24 @@ jobs:
SGLANG_USE_MODELSCOPE: true SGLANG_USE_MODELSCOPE: true
SGLANG_IS_IN_CI: true SGLANG_IS_IN_CI: true
HF_ENDPOINT: https://hf-mirror.com HF_ENDPOINT: https://hf-mirror.com
TORCH_EXTENSIONS_DIR: /tmp/torch_extensions TORCH_EXTENSIONS_DIR: /tmp/torch_extensions
PYTORCH_NPU_ALLOC_CONF: "expandable_segments:True" PYTORCH_NPU_ALLOC_CONF: "expandable_segments:True"
STREAMS_PER_DEVICE: 32 STREAMS_PER_DEVICE: 32
SGLANG_DIFFUSION_ARTIFACT_DIR: ${{ github.workspace }}/diffusion-failures SGLANG_DIFFUSION_ARTIFACT_DIR: ${{ github.workspace }}/diffusion-failures
run: | run: |
cd python cd python
python3 sglang/multimodal_gen/test/run_suite_npu.py --suite 8-npu python3 sglang/multimodal_gen/test/run_suite_npu.py --suite 8-npu
- name: Upload diffusion failure artifacts - name: Upload diffusion failure artifacts
if: always() if: always()
uses: actions/upload-artifact@v4 uses: actions/upload-artifact@v4
with: with:
name: diffusion-failures-npu-8-${{ github.run_attempt }} name: diffusion-failures-npu-8-${{ github.run_attempt }}
path: diffusion-failures/ path: diffusion-failures/
if-no-files-found: ignore if-no-files-found: ignore
retention-days: 7 retention-days: 7
pr-test-npu-finish: pr-test-npu-finish:
needs: needs:
[ [
check-changes, check-changes,
scripts/ci/utils/install_rustup.sh
+24 -9
View File
@@ -4,8 +4,12 @@
# setuptools-rust. Minimum supported version is 1.85 (edition 2024). # setuptools-rust. Minimum supported version is 1.85 (edition 2024).
set -euxo pipefail set -euxo pipefail
# Pick up cargo if rustup was installed in a previous CI step. # Make cargo/rustc visible to the rest of this shell and to subsequent
# GitHub Actions steps in the same job.
export PATH="${CARGO_HOME:-$HOME/.cargo}/bin:${PATH}" export PATH="${CARGO_HOME:-$HOME/.cargo}/bin:${PATH}"
if [ -n "${GITHUB_PATH:-}" ]; then
echo "${CARGO_HOME:-$HOME/.cargo}/bin" >> "${GITHUB_PATH}"
fi
if command -v cargo >/dev/null 2>&1 && command -v rustc >/dev/null 2>&1; then if command -v cargo >/dev/null 2>&1 && command -v rustc >/dev/null 2>&1; then
echo "rust already installed: $(rustc --version), $(cargo --version)" echo "rust already installed: $(rustc --version), $(cargo --version)"
@@ -27,14 +31,25 @@ if ! command -v curl >/dev/null 2>&1; then
fi fi
fi fi
curl --proto '=https' --tlsv1.2 --retry 3 --retry-delay 2 -sSf https://sh.rustup.rs \ if [ -n "${RUSTUP_CACHE_URL:-}" ]; then
| sh -s -- -y --no-modify-path # An in-cluster HTTP mirror is available (e.g. on NPU runners).
export RUSTUP_DIST_SERVER="${RUSTUP_CACHE_URL}/rustup"
# Make cargo/rustc visible to the rest of this shell and to subsequent export RUSTUP_UPDATE_ROOT="${RUSTUP_CACHE_URL}/rustup/rustup"
# GitHub Actions steps in the same job. case "$(uname -m)" in
export PATH="${CARGO_HOME:-$HOME/.cargo}/bin:${PATH}" x86_64) RUSTUP_ARCH="x86_64-unknown-linux-gnu" ;;
if [ -n "${GITHUB_PATH:-}" ]; then aarch64) RUSTUP_ARCH="aarch64-unknown-linux-gnu" ;;
echo "${CARGO_HOME:-$HOME/.cargo}/bin" >> "${GITHUB_PATH}" *) echo "ERROR: unsupported arch $(uname -m)"; exit 1 ;;
esac
RUSTUP_TMP="$(mktemp -d)"
trap 'rm -rf "${RUSTUP_TMP}"' EXIT
curl --retry 3 --retry-delay 2 -sSfL \
"${RUSTUP_UPDATE_ROOT}/dist/${RUSTUP_ARCH}/rustup-init" \
-o "${RUSTUP_TMP}/rustup-init"
chmod +x "${RUSTUP_TMP}/rustup-init"
"${RUSTUP_TMP}/rustup-init" -y --no-modify-path
else
curl --proto '=https' --tlsv1.2 --retry 3 --retry-delay 2 -sSf https://sh.rustup.rs \
| sh -s -- -y --no-modify-path
fi fi
rustc --version rustc --version