vllm-ascend 0.9.0rc2__tar.gz → 0.9.2rc1__tar.gz

{vllm_ascend-0.9.0rc2 → vllm_ascend-0.9.2rc1}/.github/Dockerfile.buildwheel

@@ -15,17 +15,16 @@
 # This file is a part of the vllm-ascend project.
 #
 ARG PY_VERSION=3.10
-FROM quay.io/ascend/cann:8.0.0-910b-ubuntu22.04-py${PY_VERSION}
+FROM quay.io/ascend/manylinux:8.0.0-910b-manylinux_2_28-py${PY_VERSION}
 
 ARG COMPILE_CUSTOM_KERNELS=1
 
 # Define environments
 ENV DEBIAN_FRONTEND=noninteractive
 ENV COMPILE_CUSTOM_KERNELS=${COMPILE_CUSTOM_KERNELS}
-RUN apt-get update -y && \
-    apt-get install -y python3-pip git vim wget net-tools gcc g++ cmake libnuma-dev && \
-    rm -rf /var/cache/apt/* && \
-    rm -rf /var/lib/apt/lists/*
+RUN yum update -y && \
+    yum install -y python3-pip git vim wget net-tools gcc gcc-c++ make cmake numactl-devel && \
+    rm -rf /var/cache/yum
 
 WORKDIR /workspace
 
@@ -41,8 +40,6 @@ RUN source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
     export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/Ascend/ascend-toolkit/latest/`uname -i`-linux/devlib && \
     cd vllm-ascend && \
     python3 setup.py bdist_wheel && \
-    ls -l dist && \
-    for f in dist/*.whl; do mv "$f" "$(echo "$f" | sed -e 's/-linux_x86_64\.whl$/-manylinux1_x86_64.whl/' -e 's/-linux_aarch64\.whl$/-manylinux2014_aarch64.whl/')"; done && \
-    ls -l dist
+    ls -l dist 
 
 CMD ["/bin/bash"]

{vllm_ascend-0.9.0rc2 → vllm_ascend-0.9.2rc1}/.github/ISSUE_TEMPLATE/110-user-story.yml

@@ -1,5 +1,5 @@
 name: 📚 User Story
-description: Apply for an user story to be displayed on https://vllm-ascend.readthedocs.org/user_stories/index.html
+description: Apply for an user story to be displayed on https://vllm-ascend.readthedocs.io/en/latest/community/user_stories/index.html
 title: "[User Story]: "
 labels: ["user-story"]
 

vllm_ascend-0.9.2rc1/.github/ISSUE_TEMPLATE/900-release-checklist.yml

@@ -0,0 +1,100 @@
+name: Release Checklist
+description: Generate a release checklist issue when prepare a new release.(Used for release team)
+title: "[Release]: Release checklist for v"
+
+body:
+- type: textarea
+  attributes:
+    description: >
+      Brief info for the new release.
+    label: Release Checklist
+    value: >
+      **Release Version**: 
+
+      **Release Branch**: 
+
+      **Release Date**: 
+
+      **Release Manager**: 
+- type: textarea
+  attributes:
+    description: >
+      Release notes.
+    label: Prepare Release Note
+    value: >
+      - [ ] Create a new issue for release feedback
+
+      - [ ] Write the release note PR.
+
+        - [ ] Update the feedback issue link in docs/source/faqs.md
+
+        - [ ] Add release note to docs/source/user_guide/release_notes.md
+
+        - [ ] Update version info in docs/source/community/versioning_policy.md
+
+        - [ ] Update contributor info in docs/source/community/contributors.md
+
+        - [ ] Update package version in docs/conf.py
+- type: textarea
+  attributes:
+    description: >
+      Make sure the code is merged.
+    label: PR need Merge
+    value: >
+      - [ ] PR link1
+
+      - [ ] PR link2
+
+      - [ ] ...
+- type: textarea
+  attributes:
+    description: >
+      Make sure the new Feature/Function is tested
+    label: Functional Test
+    value: >
+      - [ ] Feature1
+
+      - [ ] Bug1
+
+      - [ ] ...
+- type: textarea
+  attributes:
+    description: >
+      Make sure the doc is updated.
+    label: Doc Test
+    value: >
+      - [ ] Tutorial is updated.
+
+      - [ ] User Guide is updated.
+
+      - [ ] Developer Guide is updated.
+- type: textarea
+  attributes:
+    description: >
+      Make sure the artifacts is ready
+    label: Prepare Artifacts
+    value: >
+      - [ ] Docker image is ready.
+
+      - [ ] Wheel package is ready.
+- type: textarea
+  attributes:
+    description: >
+      Start to release.
+    label: Release Step
+    value: >
+      - [ ] Release note PR is merged.
+
+      - [ ] Post the release on GitHub release page.
+
+      - [ ] Generate official doc page on https://app.readthedocs.org/dashboard/
+
+      - [ ] Wait for the wheel package to be available on https://pypi.org/project/vllm-ascend
+
+      - [ ] Wait for the docker image to be available on https://quay.io/ascend/vllm-ascend
+
+      - [ ] Upload 310p wheel to Github release page
+
+      - [ ] Broadcast the release news (By message, blog , etc)
+
+      - [ ] Close this issue

vllm_ascend-0.9.2rc1/.github/format_pr_body.sh

@@ -0,0 +1,56 @@
+#
+# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# This file is a part of the vllm-ascend project.
+# Adapted from vllm/.github/scripts/cleanup_pr_body.sh
+
+#!/bin/bash
+
+set -eux
+
+# ensure 2 argument is passed
+if [ "$#" -ne 3 ]; then
+    echo "Usage: $0 <pr_number> <vllm_version> <vllm_commit>"
+    exit 1
+fi
+
+PR_NUMBER=$1
+VLLM_VERSION=$2
+VLLM_COMMIT=$3
+OLD=/tmp/orig_pr_body.txt
+NEW=/tmp/new_pr_body.txt
+
+gh pr view --json body --template "{{.body}}" "${PR_NUMBER}" > "${OLD}"
+cp "${OLD}" "${NEW}"
+
+# Remove notes in pr description and add vLLM version and commit
+sed -i '/<!--/,/-->/d' "${NEW}"
+sed -i '/- vLLM .*$/d' "${NEW}"
+{
+    echo ""
+    echo "- vLLM version: $VLLM_VERSION"
+    echo "- vLLM main: $VLLM_COMMIT"
+    echo ""
+} >> "${NEW}"
+
+# Run this only if ${NEW} is different than ${OLD}
+if ! cmp -s "${OLD}" "${NEW}"; then
+    echo
+    echo "Updating PR body:"
+    echo
+    cat "${NEW}"
+    gh pr edit --body-file "${NEW}" "${PR_NUMBER}"
+else
+    echo "No changes needed"
+fi

{vllm_ascend-0.9.0rc2 → vllm_ascend-0.9.2rc1}/.github/workflows/accuracy_test.yaml

@@ -22,6 +22,9 @@
 name: Benchmarks / accuracy
 
 on:
+  schedule:
+    # Runs every 6 hours
+    - cron:  '0 */6 * * *'
   pull_request:
     types: [ labeled ]
   workflow_dispatch:
@@ -34,8 +37,8 @@ on:
         # Current supported vLLM versions
         options:
           - main
-          - v0.9.0.1
-          - v0.9.0
+          - v0.9.2
+          - v0.9.1
           - v0.7.3
       vllm-ascend-version:
         description: 'vllm-ascend version:'
@@ -43,6 +46,7 @@ on:
         type: choice
         options:
           - main
+          - v0.9.1-dev
           - v0.7.3-dev
       models:
         description: 'model:'
@@ -50,9 +54,9 @@ on:
         type: choice
         options:
           - all
-          - Qwen/Qwen2.5-7B-Instruct
           - Qwen/Qwen2.5-VL-7B-Instruct
           - Qwen/Qwen3-8B-Base
+          - Qwen/Qwen3-30B-A3B
         default: 'all'
 
 # Bash shells do not use ~/.profile or ~/.bashrc so these shells need to be explicitly
@@ -74,56 +78,56 @@ jobs:
       ${{
       (contains(github.event.pull_request.labels.*.name, 'accuracy-test') ||
       contains(github.event.pull_request.labels.*.name, 'vl-accuracy-test') ||
+      contains(github.event.pull_request.labels.*.name, 'moe-accuracy-test') ||
       contains(github.event.pull_request.labels.*.name, 'dense-accuracy-test')) &&
       contains(github.event.pull_request.labels.*.name, 'ready-for-test') ||
-      github.event_name == 'workflow_dispatch'
+      github.event_name == 'workflow_dispatch' || github.event_name == 'schedule'
       }}
     runs-on: >-
       ${{
-          (matrix.model_name == 'Qwen/Qwen2.5-VL-7B-Instruct' && 'linux-arm64-npu-4') ||
+          (matrix.model_name == 'Qwen/Qwen3-30B-A3B' && 'linux-arm64-npu-4') ||
           'linux-arm64-npu-2'
       }}
     strategy:
       matrix:
-        vllm_use_version: [0, 1]
         # the accuracy test will run:
         # 1. workflow_dispatch with models input
-        #   - all: Qwen/Qwen2.5-7B-Instruct, Qwen/Qwen2.5-VL-7B-Instruct, Qwen/Qwen3-8B-Base
-        #   - specified but not all: Qwen/Qwen2.5-7B-Instruct, Qwen/Qwen2.5-VL-7B-Instruct, Qwen/Qwen3-8B-Base
+        #   - all: Qwen/Qwen3-30B-A3B, Qwen/Qwen2.5-VL-7B-Instruct, Qwen/Qwen3-8B-Base
+        #   - specified but not all: Qwen/Qwen3-30B-A3B, Qwen/Qwen2.5-VL-7B-Instruct, Qwen/Qwen3-8B-Base
         # 2. PR labeled with "*-accuracy-test"
-        #   - accuracy-test: Qwen/Qwen2.5-7B-Instruct, Qwen/Qwen2.5-VL-7B-Instruct
-        #   - dense-accuracy-test: Qwen/Qwen2.5-7B-Instruct
+        #   - accuracy-test: Qwen/Qwen3-8B-Base, Qwen/Qwen2.5-VL-7B-Instruct, Qwen/Qwen3-30B-A3B
+        #   - dense-accuracy-test: Qwen/Qwen3-8B-Base
         #   - vl-accuracy-test: Qwen/Qwen2.5-VL-7B-Instruct
+        #   - moe-accuracy-test: Qwen/Qwen3-30B-A3B
         model_name: ${{ fromJSON(
+          (github.event_name == 'schedule' &&
+            '["Qwen/Qwen3-30B-A3B","Qwen/Qwen2.5-VL-7B-Instruct","Qwen/Qwen3-8B-Base"]') ||
           (github.event.inputs.models == 'all' &&
-            '["Qwen/Qwen2.5-7B-Instruct","Qwen/Qwen2.5-VL-7B-Instruct","Qwen/Qwen3-8B-Base"]') ||
-          (github.event.inputs.models == 'Qwen/Qwen2.5-7B-Instruct' &&
-            '["Qwen/Qwen2.5-7B-Instruct"]') ||
+            '["Qwen/Qwen3-30B-A3B","Qwen/Qwen2.5-VL-7B-Instruct","Qwen/Qwen3-8B-Base"]') ||
+          (github.event.inputs.models == 'Qwen/Qwen3-30B-A3B' &&
+            '["Qwen/Qwen3-30B-A3B"]') ||
           (github.event.inputs.models == 'Qwen/Qwen2.5-VL-7B-Instruct' &&
             '["Qwen/Qwen2.5-VL-7B-Instruct"]') ||
           (github.event.inputs.models == 'Qwen/Qwen3-8B-Base' &&
             '["Qwen/Qwen3-8B-Base"]') ||
           contains(github.event.pull_request.labels.*.name, 'accuracy-test') &&
-            '["Qwen/Qwen2.5-7B-Instruct","Qwen/Qwen2.5-VL-7B-Instruct"]' ||
+            '["Qwen/Qwen3-8B-Base","Qwen/Qwen2.5-VL-7B-Instruct", "Qwen/Qwen3-30B-A3B"]' ||
           contains(github.event.pull_request.labels.*.name, 'dense-accuracy-test') &&
-            '["Qwen/Qwen2.5-7B-Instruct"]' ||
+            '["Qwen/Qwen3-8B-Base"]' ||
           contains(github.event.pull_request.labels.*.name, 'vl-accuracy-test') &&
-            '["Qwen/Qwen2.5-VL-7B-Instruct"]'
+            '["Qwen/Qwen2.5-VL-7B-Instruct"]' ||
+          contains(github.event.pull_request.labels.*.name, 'moe-accuracy-test') &&
+            '["Qwen/Qwen3-30B-A3B"]'
          ) }}
-        # Remove exclude after https://github.com/vllm-project/vllm-ascend/issues/1044 resolved
-        exclude:
-          - model_name: Qwen/Qwen2.5-VL-7B-Instruct
-            vllm_use_version: 1
 
       fail-fast: false
-    name: ${{ matrix.model_name }} accuracy V${{ matrix.vllm_use_version }}
+    name: ${{ matrix.model_name }} accuracy
     container:
       image: m.daocloud.io/quay.io/ascend/cann:8.1.rc1-910b-ubuntu22.04-py3.10
       env:
-        HF_ENDPOINT: https://hf-mirror.com
-        HF_TOKEN: ${{ secrets.HF_TOKEN }}
         DATASET_SOURCE: ModelScope
         VLLM_USE_MODELSCOPE: True
+        USE_MODELSCOPE_HUB: 1
         # 1. If version specified (work_dispatch), do specified branch accuracy test
         # 2. If no version (labeled PR), do accuracy test by default ref:
         # The branch, tag or SHA to checkout. When checking out the repository that
@@ -159,7 +163,7 @@ jobs:
           repository: vllm-project/vllm
           path: ./vllm-empty
           # Please also update this when bump matched version
-          ref: ${{ github.event.inputs.vllm-version || 'v0.9.0' }}
+          ref: ${{ github.event.inputs.vllm-version || 'v0.9.2' }}
 
       - name: Install vllm-project/vllm from source
         working-directory: ./vllm-empty
@@ -174,13 +178,32 @@ jobs:
 
       - name: Install vllm-project/vllm-ascend
         working-directory: ./vllm-ascend
+        env:
+          PIP_EXTRA_INDEX_URL: https://mirrors.huaweicloud.com/ascend/repos/pypi
         run: |
           pip install -r requirements-dev.txt
-          pip install -e .
+          pip install -v -e . 
+            
+      - name: Get vLLM commit hash and URL
+        working-directory: ./vllm-empty
+        run: |
+          VLLM_COMMIT=$(git rev-parse --short=7 HEAD)
+          echo "VLLM_COMMIT=$VLLM_COMMIT" >> $GITHUB_ENV
+
+      - name: Get vLLM-Ascend commit hash and URL
+        working-directory: ./vllm-ascend
+        run: |
+          VLLM_ASCEND_COMMIT=$(git rev-parse --short=7 HEAD)
+          echo "VLLM_ASCEND_COMMIT=$VLLM_ASCEND_COMMIT" >> $GITHUB_ENV
+
+      - name: Print resolved hashes
+        run: |
+          echo "vLLM       : ${{ env.VLLM_COMMIT }}"
+          echo "vLLM-Ascend: ${{ env.VLLM_ASCEND_COMMIT }}"
 
       - name: Install lm-eval, ray, and datasets
         run: |
-            pip install lm-eval
+            pip install lm-eval==0.4.8
 
       - name: Collect version info
         run: |
@@ -212,15 +235,14 @@ jobs:
           echo "vLLM: ${{ env.GHA_VLLM_VERSION }}"
           echo "vLLM Ascend: ${{ env.GHA_VLLM_ASCEND_VERSION }}"
 
-      - name: Run Accuracy Test for V${{ matrix.vllm_use_version }}
+      - name: Run Accuracy Test
         id: report
         working-directory: ./benchmarks
         env:
           PYTORCH_NPU_ALLOC_CONF: max_split_size_mb:256
-          VLLM_USE_V1: ${{ matrix.vllm_use_version }}
         run: |
           model_base_name=$(basename ${{ matrix.model_name }})
-          markdown_name="${model_base_name}-V${{ matrix.vllm_use_version }}"
+          markdown_name="${model_base_name}"
           echo "markdown_name=$markdown_name"
           echo "markdown_name=$markdown_name" >> $GITHUB_OUTPUT
           mkdir -p ./accuracy
@@ -232,7 +254,9 @@ jobs:
             --cann_version "${{ env.GHA_CANN_VERSION }}" \
             --torch_npu_version "${{ env.GHA_TORCH_NPU_VERSION }}" \
             --torch_version "${{ env.GHA_TORCH_VERSION }}" \
-            --vllm_version "${{ env.GHA_VLLM_VERSION }}"
+            --vllm_version "${{ env.GHA_VLLM_VERSION }}" \
+            --vllm_commit "${{ env.VLLM_COMMIT }}" \
+            --vllm_ascend_commit "${{ env.VLLM_ASCEND_COMMIT }}" \
 
       - name: Generate step summary
         if: ${{ always() }}
@@ -244,12 +268,122 @@ jobs:
           SAFE_VLLM_ASCEND_VERSION="${GHA_VLLM_ASCEND_VERSION//\//-}"
           echo "SAFE_VLLM_ASCEND_VERSION=$SAFE_VLLM_ASCEND_VERSION" >> "$GITHUB_ENV"
 
-      - name: Upload Report for V${{ matrix.vllm_use_version }}
-        if: ${{ github.event_name == 'workflow_dispatch' }}
+      - name: Check report first line for failure
+        id: check_report
+        run: |
+          REPORT_PATH="./benchmarks/accuracy/${{ steps.report.outputs.markdown_name }}.md"
+          echo "Scanning $REPORT_PATH for ❌ …"
+          if grep -q '❌' "$REPORT_PATH"; then
+            echo "contains_fail=true" >> $GITHUB_OUTPUT
+          else
+            echo "contains_fail=false" >> $GITHUB_OUTPUT
+          fi
+
+      - name: Upload Report 
+        if: ${{ github.event_name == 'workflow_dispatch' && steps.check_report.outputs.contains_fail == 'false' }}
         uses: actions/upload-artifact@v4
         with:
-          name: "${{ env.SAFE_VLLM_ASCEND_VERSION }}-${{ steps.report.outputs.markdown_name }}-report"
+          name: "report-${{ env.SAFE_VLLM_ASCEND_VERSION }}-${{ steps.report.outputs.markdown_name }}"
           path: ./benchmarks/accuracy/${{ steps.report.outputs.markdown_name }}.md
           if-no-files-found: warn
           retention-days: 90
           overwrite: true
+
+  create_pr:
+    runs-on: ubuntu-latest
+    needs: accuracy_tests
+    if: ${{ github.event_name == 'workflow_dispatch' }}
+    env:
+      UPSTREAM_REPO: vllm-project/vllm-ascend
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+        with:
+          repository: vllm-ascend-ci/vllm-ascend
+          token: ${{ secrets.PAT_TOKEN }}
+          ref: main
+
+      - name: Add upstream remote
+        run: |
+          git remote add upstream https://github.com/${{ env.UPSTREAM_REPO }}.git
+          git fetch upstream
+          git remote -v
+
+      - name: Set Git user info dynamically
+        run: |
+          git config user.name "${{ github.actor }}"
+          git config user.email "${{ github.actor }}@users.noreply.github.com"
+
+      - name: Create or switch to branch
+        run: |
+          TIMESTAMP=$(date +%Y%m%d%H%M%S)
+          BRANCH_NAME="auto-pr/accuracy-report-${TIMESTAMP}"
+          echo "BRANCH_NAME=${BRANCH_NAME}" >> $GITHUB_ENV
+          git checkout -B "${BRANCH_NAME}" upstream/${{ github.event.inputs.vllm-ascend-version }}
+
+      - name: Download only current run reports
+        uses: actions/download-artifact@v4
+        with:
+          path: ./docs/source/developer_guide/evaluation/accuracy_report
+          pattern: report-*
+          github-token: ${{ secrets.GITHUB_TOKEN }}
+          run-id: ${{ github.run_id }}
+
+      - name: Delete old report
+        run: |
+          find ./docs/source/developer_guide/evaluation/accuracy_report -maxdepth 1 -type f -name '*.md' ! -name 'index.md' -delete
+          find ./docs/source/developer_guide/evaluation/accuracy_report -mindepth 2 -type f -name '*.md' -exec mv -f {} ./docs/source/developer_guide/evaluation/accuracy_report \;
+          find ./docs/source/developer_guide/evaluation/accuracy_report -mindepth 1 -type d -empty -delete
+
+      - name: Update accuracy_report/index.md
+        run: |
+          REPORT_DIR="./docs/source/developer_guide/evaluation/accuracy_report"
+          INDEX_MD="$REPORT_DIR/index.md"
+          {
+            echo "# Accuracy Report"
+            echo ""
+            echo ":::{toctree}"
+            echo ":caption: Accuracy Report"
+            echo ":maxdepth: 1"
+            
+            for report in "$REPORT_DIR"/*.md; do
+              filename="$(basename "$report" .md)"
+              if [ "$filename" != "index" ]; then
+                echo "$filename"
+              fi
+            done
+            echo ":::"
+          } > "$INDEX_MD"
+
+      - name: push accuracy report
+        env:
+          GITHUB_TOKEN: ${{ secrets.PAT_TOKEN }}
+        run: |
+          git add ./docs/source/developer_guide/evaluation/accuracy_report/*.md
+          git commit -s -m "[Doc] Update accuracy reports for ${{ github.event.inputs.vllm-ascend-version }}"
+          git push -f origin "${{ env.BRANCH_NAME }}"
+
+      - name: Create PR in upstream via API
+        uses: actions/github-script@v7
+        with:
+          github-token: ${{ secrets.PAT_TOKEN }}
+          script: |
+            const pr = await github.rest.pulls.create({
+              owner: 'vllm-project',
+              repo: 'vllm-ascend',
+              head: `vllm-ascend-ci:${{ env.BRANCH_NAME }}`,
+              base: '${{ github.event.inputs.vllm-ascend-version }}',
+              title: `[Doc] Update accuracy reports for ${{ github.event.inputs.vllm-ascend-version }}`,
+              body: `The accuracy results running on NPU Altlas A2 have changed, updating reports for:
+            ${{ 
+              github.event.inputs.models == 'all' 
+                && 'All models (Qwen/Qwen3-30B-A3B, Qwen2.5-VL-7B-Instruct, Qwen3-8B-Base)' 
+                || github.event.inputs.models 
+            }}
+            
+            - [Workflow run][1]
+            
+            [1]: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}`
+            });
+            core.info(`Created PR #${pr.data.number}`);
+ 

vllm_ascend-0.9.2rc1/.github/workflows/format_pr_body.yaml

@@ -0,0 +1,63 @@
+#
+# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# This file is a part of the vllm-ascend project.
+#
+
+name: format / pr body
+
+on:
+  # The PR updated when PR opened and push new commits
+  pull_request_target:
+    types: [opened, synchronize]
+    branches:
+      - 'main'
+
+permissions:
+  pull-requests: write
+
+jobs:
+  update-description:
+    name: update vLLM version
+    runs-on: ubuntu-latest
+
+    steps:
+      - name: Checkout vllm-project/vllm repo
+        uses: actions/checkout@v4
+        with:
+          repository: vllm-project/vllm
+          path: ./vllm-empty
+
+      - name: Get vLLM version
+        working-directory: ./vllm-empty
+        run: |
+          VLLM_COMMIT=$(git rev-parse HEAD)
+          echo "VLLM_COMMIT=https://github.com/vllm-project/vllm/commit/$VLLM_COMMIT" >> $GITHUB_ENV
+
+      - name: Checkout repository
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+
+      - name: Set up Python
+        uses: actions/setup-python@42375524e23c412d93fb67b49958b491fce71c38 # v5.4.0
+
+      - name: Get vLLM release version
+        run: |
+          VLLM_VERSION=$(python3 docs/source/conf.py | jq .ci_vllm_version | tr -d '"')
+          echo "VLLM_VERSION=$VLLM_VERSION" >> $GITHUB_ENV
+
+      - name: Update PR description
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          bash .github/format_pr_body.sh "${{ github.event.number }}" "${{ env.VLLM_VERSION }}" "${{ env.VLLM_COMMIT }}"