From 66ce2b4f18e2d73e3d66a86ea825f23887c80885 Mon Sep 17 00:00:00 2001 From: bingxche Date: Thu, 30 Apr 2026 02:44:09 -0500 Subject: [PATCH 1/7] [AMD] Add DeepSeek-V4 ROCm7.2 nightly Docker release Made-with: Cursor --- ...docker-amd-rocm720-nightly-deepseek-v4.yml | 98 +++++++++++++++++++ 1 file changed, 98 insertions(+) create mode 100644 .github/workflows/release-docker-amd-rocm720-nightly-deepseek-v4.yml diff --git a/.github/workflows/release-docker-amd-rocm720-nightly-deepseek-v4.yml b/.github/workflows/release-docker-amd-rocm720-nightly-deepseek-v4.yml new file mode 100644 index 000000000000..5a65d45e10ae --- /dev/null +++ b/.github/workflows/release-docker-amd-rocm720-nightly-deepseek-v4.yml @@ -0,0 +1,98 @@ +name: Release Docker Images Nightly ROCm7.2 (AMD) DeepSeek-V4 + +on: + workflow_dispatch: + schedule: + - cron: '0 12 * * *' + +concurrency: + group: ${{ github.workflow }}-${{ github.event.number || github.sha }} + cancel-in-progress: true + +jobs: + publish: + if: github.repository == 'sgl-project/sglang' + runs-on: amd-docker-scale + environment: 'prod' + strategy: + fail-fast: false + matrix: + gpu_arch: ['gfx942-rocm720', 'gfx950-rocm720'] + build_type: ['all'] + steps: + - name: Checkout repository + uses: actions/checkout@v4 + with: + ref: amd/deepseek_v4 + fetch-depth: 0 # Required for git describe to find tags + + - name: Set up Python + uses: actions/setup-python@v5 + with: + python-version: "3.10" + + - name: "Set Date" + run: | + echo "DATE=$(date +%Y%m%d)" >> $GITHUB_ENV + + - name: Get version from latest tag + id: version + run: | + # Use the shared helper so stable/post releases sort above rc tags. + VERSION=$(python3 python/tools/get_version_tag.py --tag-only | sed 's/^v//') + + if [ -z "$VERSION" ]; then + echo "::error::Could not determine version from git tags" + exit 1 + fi + + # Get short commit hash of current HEAD + COMMIT_SHA=$(git rev-parse HEAD) + COMMIT_HASH=${COMMIT_SHA:0:7} + + # Compose pretend version for setuptools_scm: e.g., 0.5.8.post1.dev20260211+g1a2b3c4 + PRETEND_VERSION="${VERSION}.dev${{ env.DATE }}+g${COMMIT_HASH}" + + echo "commit_sha=${COMMIT_SHA}" >> "$GITHUB_OUTPUT" + echo "commit_hash=${COMMIT_HASH}" >> "$GITHUB_OUTPUT" + echo "version=${VERSION}" >> "$GITHUB_OUTPUT" + echo "pretend_version=${PRETEND_VERSION}" >> "$GITHUB_OUTPUT" + echo "DeepSeek V4 commit: ${COMMIT_SHA}" + echo "Detected version: ${VERSION}" + echo "Pretend version for pip: ${PRETEND_VERSION}" + + - name: Login to Docker Hub + uses: docker/login-action@v2 + with: + username: ${{ secrets.DOCKERHUB_AMD_USERNAME }} + password: ${{ secrets.DOCKERHUB_AMD_TOKEN }} + + - name: Build and Push DSv4 image to rocm/sgl-dev + run: | + version=${{ steps.version.outputs.version }} + pretend_version=${{ steps.version.outputs.pretend_version }} + echo "Version: ${version}" + echo "Pretend version: ${pretend_version}" + + if [ "${{ matrix.gpu_arch }}" = "gfx942-rocm720" ]; then + rocm_tag="rocm720-mi30x" + elif [ "${{ matrix.gpu_arch }}" = "gfx950-rocm720" ]; then + rocm_tag="rocm720-mi35x" + else + echo "Unsupported gfx arch" + exit 1 + fi + + image_tag="${rocm_tag}-${{ steps.version.outputs.commit_hash }}-${{ env.DATE }}-DSv4" + echo "IMAGE_TAG=${image_tag}" >> "$GITHUB_ENV" + echo "Building rocm/sgl-dev:${image_tag} from amd/deepseek_v4 @ ${{ steps.version.outputs.commit_sha }}" + + docker build . -f docker/rocm.Dockerfile \ + --build-arg SGL_BRANCH=${{ steps.version.outputs.commit_sha }} \ + --build-arg BUILD_TYPE=${{ matrix.build_type }} \ + --build-arg GPU_ARCH=${{ matrix.gpu_arch }} \ + --build-arg ENABLE_MORI=1 \ + --build-arg SETUPTOOLS_SCM_PRETEND_VERSION=${pretend_version} \ + -t rocm/sgl-dev:${image_tag} \ + --no-cache + docker push rocm/sgl-dev:${image_tag} From ff1861c7d2a98954e90168adcb9ac6cf9c114b48 Mon Sep 17 00:00:00 2001 From: bingxche Date: Thu, 30 Apr 2026 02:47:00 -0500 Subject: [PATCH 2/7] use mi300 8 gpu runner to trail fast --- .../release-docker-amd-rocm720-nightly-deepseek-v4.yml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/.github/workflows/release-docker-amd-rocm720-nightly-deepseek-v4.yml b/.github/workflows/release-docker-amd-rocm720-nightly-deepseek-v4.yml index 5a65d45e10ae..52ba05ca968d 100644 --- a/.github/workflows/release-docker-amd-rocm720-nightly-deepseek-v4.yml +++ b/.github/workflows/release-docker-amd-rocm720-nightly-deepseek-v4.yml @@ -12,7 +12,7 @@ concurrency: jobs: publish: if: github.repository == 'sgl-project/sglang' - runs-on: amd-docker-scale + runs-on: linux-mi300-8gpu-sglang environment: 'prod' strategy: fail-fast: false From 8d9095113c1fbb808f3f2516c5d88232535fe92b Mon Sep 17 00:00:00 2001 From: bingxche Date: Thu, 30 Apr 2026 02:52:34 -0500 Subject: [PATCH 3/7] [AMD] Add DSv4 ROCm7.2 release trial job Made-with: Cursor --- .../release-docker-amd-rocm720-nightly.yml | 87 +++++++++++++++++++ 1 file changed, 87 insertions(+) diff --git a/.github/workflows/release-docker-amd-rocm720-nightly.yml b/.github/workflows/release-docker-amd-rocm720-nightly.yml index d801c77cc287..e5675479e9de 100644 --- a/.github/workflows/release-docker-amd-rocm720-nightly.yml +++ b/.github/workflows/release-docker-amd-rocm720-nightly.yml @@ -160,3 +160,90 @@ jobs: docker pull "${src}" docker tag "${src}" "${dst}" docker push "${dst}" + + publish_dsv4: + if: github.repository == 'sgl-project/sglang' + runs-on: linux-mi300-8gpu-sglang + environment: 'prod' + strategy: + fail-fast: false + matrix: + gpu_arch: ['gfx942-rocm720', 'gfx950-rocm720'] + build_type: ['all'] + steps: + - name: Checkout repository + uses: actions/checkout@v4 + with: + ref: amd/deepseek_v4 + fetch-depth: 0 # Required for git describe to find tags + + - name: Set up Python + uses: actions/setup-python@v5 + with: + python-version: "3.10" + + - name: "Set Date" + run: | + echo "DATE=$(date +%Y%m%d)" >> $GITHUB_ENV + + - name: Get version from latest tag + id: version + run: | + # Use the shared helper so stable/post releases sort above rc tags. + VERSION=$(python3 python/tools/get_version_tag.py --tag-only | sed 's/^v//') + + if [ -z "$VERSION" ]; then + echo "::error::Could not determine version from git tags" + exit 1 + fi + + # Get short commit hash of current HEAD + COMMIT_SHA=$(git rev-parse HEAD) + COMMIT_HASH=${COMMIT_SHA:0:7} + + # Compose pretend version for setuptools_scm: e.g., 0.5.8.post1.dev20260211+g1a2b3c4 + PRETEND_VERSION="${VERSION}.dev${{ env.DATE }}+g${COMMIT_HASH}" + + echo "commit_sha=${COMMIT_SHA}" >> "$GITHUB_OUTPUT" + echo "commit_hash=${COMMIT_HASH}" >> "$GITHUB_OUTPUT" + echo "version=${VERSION}" >> "$GITHUB_OUTPUT" + echo "pretend_version=${PRETEND_VERSION}" >> "$GITHUB_OUTPUT" + echo "DeepSeek V4 commit: ${COMMIT_SHA}" + echo "Detected version: ${VERSION}" + echo "Pretend version for pip: ${PRETEND_VERSION}" + + - name: Login to Docker Hub + uses: docker/login-action@v2 + with: + username: ${{ secrets.DOCKERHUB_AMD_USERNAME }} + password: ${{ secrets.DOCKERHUB_AMD_TOKEN }} + + - name: Build and Push DSv4 image to rocm/sgl-dev + run: | + version=${{ steps.version.outputs.version }} + pretend_version=${{ steps.version.outputs.pretend_version }} + echo "Version: ${version}" + echo "Pretend version: ${pretend_version}" + + if [ "${{ matrix.gpu_arch }}" = "gfx942-rocm720" ]; then + rocm_tag="rocm720-mi30x" + elif [ "${{ matrix.gpu_arch }}" = "gfx950-rocm720" ]; then + rocm_tag="rocm720-mi35x" + else + echo "Unsupported gfx arch" + exit 1 + fi + + image_tag="${rocm_tag}-${{ steps.version.outputs.commit_hash }}-${{ env.DATE }}-DSv4" + echo "IMAGE_TAG=${image_tag}" >> "$GITHUB_ENV" + echo "Building rocm/sgl-dev:${image_tag} from amd/deepseek_v4 @ ${{ steps.version.outputs.commit_sha }}" + + docker build . -f docker/rocm.Dockerfile \ + --build-arg SGL_BRANCH=${{ steps.version.outputs.commit_sha }} \ + --build-arg BUILD_TYPE=${{ matrix.build_type }} \ + --build-arg GPU_ARCH=${{ matrix.gpu_arch }} \ + --build-arg ENABLE_MORI=1 \ + --build-arg SETUPTOOLS_SCM_PRETEND_VERSION=${pretend_version} \ + -t rocm/sgl-dev:${image_tag} \ + --no-cache + docker push rocm/sgl-dev:${image_tag} From 0ea5242b6446ec0d475073f9f3bcf039d63aac08 Mon Sep 17 00:00:00 2001 From: bingxche Date: Thu, 30 Apr 2026 02:54:26 -0500 Subject: [PATCH 4/7] [AMD] Add ROCm7.2 release job selector Made-with: Cursor --- .../release-docker-amd-rocm720-nightly.yml | 16 +++++++++++++--- 1 file changed, 13 insertions(+), 3 deletions(-) diff --git a/.github/workflows/release-docker-amd-rocm720-nightly.yml b/.github/workflows/release-docker-amd-rocm720-nightly.yml index e5675479e9de..0458a9ccd018 100644 --- a/.github/workflows/release-docker-amd-rocm720-nightly.yml +++ b/.github/workflows/release-docker-amd-rocm720-nightly.yml @@ -1,6 +1,16 @@ name: Release Docker Images Nightly ROCm7.2 (AMD) on: workflow_dispatch: + inputs: + job_select: + description: 'Select which release job to run' + required: false + type: choice + default: 'all' + options: + - 'all' + - publish + - publish_dsv4 schedule: - cron: '0 12 * * *' @@ -14,7 +24,7 @@ concurrency: jobs: publish: - if: github.repository == 'sgl-project/sglang' + if: github.repository == 'sgl-project/sglang' && (github.event_name != 'workflow_dispatch' || inputs.job_select == 'all' || inputs.job_select == 'publish') runs-on: amd-docker-scale environment: 'prod' strategy: @@ -123,7 +133,7 @@ jobs: # of publish failed; legs without an artifact will fail at download and be # the only ones marked red. push_local_registry: - if: ${{ !cancelled() && github.repository == 'sgl-project/sglang' }} + if: ${{ !cancelled() && github.repository == 'sgl-project/sglang' && (github.event_name != 'workflow_dispatch' || inputs.job_select == 'all' || inputs.job_select == 'publish') }} runs-on: linux-mi300-1gpu-sglang environment: 'prod' needs: publish @@ -162,7 +172,7 @@ jobs: docker push "${dst}" publish_dsv4: - if: github.repository == 'sgl-project/sglang' + if: github.repository == 'sgl-project/sglang' && (github.event_name != 'workflow_dispatch' || inputs.job_select == 'all' || inputs.job_select == 'publish_dsv4') runs-on: linux-mi300-8gpu-sglang environment: 'prod' strategy: From f4053e796a092eb20d515a7f8a2ac6b511481ad9 Mon Sep 17 00:00:00 2001 From: bingxche Date: Thu, 30 Apr 2026 10:05:36 -0500 Subject: [PATCH 5/7] delete release-docker-amd-rocm720-nightly-deepseek-v4.yml --- ...docker-amd-rocm720-nightly-deepseek-v4.yml | 98 ------------------- human-eval | 1 + 2 files changed, 1 insertion(+), 98 deletions(-) delete mode 100644 .github/workflows/release-docker-amd-rocm720-nightly-deepseek-v4.yml create mode 160000 human-eval diff --git a/.github/workflows/release-docker-amd-rocm720-nightly-deepseek-v4.yml b/.github/workflows/release-docker-amd-rocm720-nightly-deepseek-v4.yml deleted file mode 100644 index 52ba05ca968d..000000000000 --- a/.github/workflows/release-docker-amd-rocm720-nightly-deepseek-v4.yml +++ /dev/null @@ -1,98 +0,0 @@ -name: Release Docker Images Nightly ROCm7.2 (AMD) DeepSeek-V4 - -on: - workflow_dispatch: - schedule: - - cron: '0 12 * * *' - -concurrency: - group: ${{ github.workflow }}-${{ github.event.number || github.sha }} - cancel-in-progress: true - -jobs: - publish: - if: github.repository == 'sgl-project/sglang' - runs-on: linux-mi300-8gpu-sglang - environment: 'prod' - strategy: - fail-fast: false - matrix: - gpu_arch: ['gfx942-rocm720', 'gfx950-rocm720'] - build_type: ['all'] - steps: - - name: Checkout repository - uses: actions/checkout@v4 - with: - ref: amd/deepseek_v4 - fetch-depth: 0 # Required for git describe to find tags - - - name: Set up Python - uses: actions/setup-python@v5 - with: - python-version: "3.10" - - - name: "Set Date" - run: | - echo "DATE=$(date +%Y%m%d)" >> $GITHUB_ENV - - - name: Get version from latest tag - id: version - run: | - # Use the shared helper so stable/post releases sort above rc tags. - VERSION=$(python3 python/tools/get_version_tag.py --tag-only | sed 's/^v//') - - if [ -z "$VERSION" ]; then - echo "::error::Could not determine version from git tags" - exit 1 - fi - - # Get short commit hash of current HEAD - COMMIT_SHA=$(git rev-parse HEAD) - COMMIT_HASH=${COMMIT_SHA:0:7} - - # Compose pretend version for setuptools_scm: e.g., 0.5.8.post1.dev20260211+g1a2b3c4 - PRETEND_VERSION="${VERSION}.dev${{ env.DATE }}+g${COMMIT_HASH}" - - echo "commit_sha=${COMMIT_SHA}" >> "$GITHUB_OUTPUT" - echo "commit_hash=${COMMIT_HASH}" >> "$GITHUB_OUTPUT" - echo "version=${VERSION}" >> "$GITHUB_OUTPUT" - echo "pretend_version=${PRETEND_VERSION}" >> "$GITHUB_OUTPUT" - echo "DeepSeek V4 commit: ${COMMIT_SHA}" - echo "Detected version: ${VERSION}" - echo "Pretend version for pip: ${PRETEND_VERSION}" - - - name: Login to Docker Hub - uses: docker/login-action@v2 - with: - username: ${{ secrets.DOCKERHUB_AMD_USERNAME }} - password: ${{ secrets.DOCKERHUB_AMD_TOKEN }} - - - name: Build and Push DSv4 image to rocm/sgl-dev - run: | - version=${{ steps.version.outputs.version }} - pretend_version=${{ steps.version.outputs.pretend_version }} - echo "Version: ${version}" - echo "Pretend version: ${pretend_version}" - - if [ "${{ matrix.gpu_arch }}" = "gfx942-rocm720" ]; then - rocm_tag="rocm720-mi30x" - elif [ "${{ matrix.gpu_arch }}" = "gfx950-rocm720" ]; then - rocm_tag="rocm720-mi35x" - else - echo "Unsupported gfx arch" - exit 1 - fi - - image_tag="${rocm_tag}-${{ steps.version.outputs.commit_hash }}-${{ env.DATE }}-DSv4" - echo "IMAGE_TAG=${image_tag}" >> "$GITHUB_ENV" - echo "Building rocm/sgl-dev:${image_tag} from amd/deepseek_v4 @ ${{ steps.version.outputs.commit_sha }}" - - docker build . -f docker/rocm.Dockerfile \ - --build-arg SGL_BRANCH=${{ steps.version.outputs.commit_sha }} \ - --build-arg BUILD_TYPE=${{ matrix.build_type }} \ - --build-arg GPU_ARCH=${{ matrix.gpu_arch }} \ - --build-arg ENABLE_MORI=1 \ - --build-arg SETUPTOOLS_SCM_PRETEND_VERSION=${pretend_version} \ - -t rocm/sgl-dev:${image_tag} \ - --no-cache - docker push rocm/sgl-dev:${image_tag} diff --git a/human-eval b/human-eval new file mode 160000 index 000000000000..9ce83afd78fc --- /dev/null +++ b/human-eval @@ -0,0 +1 @@ +Subproject commit 9ce83afd78fc52ca3b0083a5a4920431d83c9ff3 From 61623dfddf66df015393d3596e448f8514399a63 Mon Sep 17 00:00:00 2001 From: bingxche Date: Thu, 30 Apr 2026 10:10:40 -0500 Subject: [PATCH 6/7] restore using amd-docker-scale --- .github/workflows/release-docker-amd-rocm720-nightly.yml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/.github/workflows/release-docker-amd-rocm720-nightly.yml b/.github/workflows/release-docker-amd-rocm720-nightly.yml index 0458a9ccd018..b42b543562b2 100644 --- a/.github/workflows/release-docker-amd-rocm720-nightly.yml +++ b/.github/workflows/release-docker-amd-rocm720-nightly.yml @@ -173,7 +173,7 @@ jobs: publish_dsv4: if: github.repository == 'sgl-project/sglang' && (github.event_name != 'workflow_dispatch' || inputs.job_select == 'all' || inputs.job_select == 'publish_dsv4') - runs-on: linux-mi300-8gpu-sglang + runs-on: amd-docker-scale environment: 'prod' strategy: fail-fast: false From b44ec5e14937739aa1b10643d25d60a5d630e70f Mon Sep 17 00:00:00 2001 From: bingxche Date: Thu, 30 Apr 2026 10:45:10 -0500 Subject: [PATCH 7/7] Remove accidental human-eval submodule Made-with: Cursor --- human-eval | 1 - 1 file changed, 1 deletion(-) delete mode 160000 human-eval diff --git a/human-eval b/human-eval deleted file mode 160000 index 9ce83afd78fc..000000000000 --- a/human-eval +++ /dev/null @@ -1 +0,0 @@ -Subproject commit 9ce83afd78fc52ca3b0083a5a4920431d83c9ff3