vllm-ascend 0.9.1rc1__tar.gz → 0.9.1rc2__tar.gz

{vllm_ascend-0.9.1rc1 → vllm_ascend-0.9.1rc2}/.github/Dockerfile.buildwheel

@@ -14,18 +14,17 @@
 # limitations under the License.
 # 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${PY_VERSION}
+ARG PY_VERSION=3.11
+FROM quay.io/ascend/manylinux:8.0.0-910b-manylinux_2_28-py${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
 
 CMD ["/bin/bash"]

{vllm_ascend-0.9.1rc1 → vllm_ascend-0.9.1rc2}/.github/dependabot.yml

@@ -2,9 +2,6 @@ version: 2
 updates:
   - package-ecosystem: "github-actions"
     directory: "/"
-    schedule:
-      # Check for updates to GitHub Actions every week
-      interval: "weekly"
     open-pull-requests-limit: 2
     reviewers:
       - "Yikun"

{vllm_ascend-0.9.1rc1 → vllm_ascend-0.9.1rc2}/.github/workflows/accuracy_test.yaml

@@ -117,7 +117,7 @@ jobs:
       fail-fast: false
     name: ${{ matrix.model_name }} accuracy V${{ matrix.vllm_use_version }}
     container:
-      image: m.daocloud.io/quay.io/ascend/cann:8.1.rc1-910b-ubuntu22.04-py3.10
+      image: swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/cann:8.2.rc1-910b-ubuntu22.04-py3.11
       env:
         HF_ENDPOINT: https://hf-mirror.com
         HF_TOKEN: ${{ secrets.HF_TOKEN }}
@@ -173,8 +173,6 @@ 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 .

{vllm_ascend-0.9.1rc1 → vllm_ascend-0.9.1rc2}/.github/workflows/image_openeuler.yml

@@ -1,4 +1,4 @@
-name: 'image / openEuler'
+name: 'image'
 # This is a docker build check and publish job:
 # 1. PR Triggered docker image build check
 #   - is for image build check
@@ -39,13 +39,9 @@ on:
 
 jobs:
   build:
-    name: vllm-ascend image build
-    runs-on: >-
-      ${{
-          github.event_name == 'push' && github.repository_owner == 'vllm-project' &&
-          'ubuntu-latest' ||
-          'ubuntu-24.04-arm'
-      }}
+    name: vllm-ascend openEuler image
+    runs-on: ubuntu-latest
+
     steps:
     - uses: actions/checkout@v4
 
@@ -94,15 +90,10 @@ jobs:
         username: ${{ vars.QUAY_USERNAME }}
         password: ${{ secrets.QUAY_PASSWORD }}
 
-    - name: Build and push 910b
+    - name: Build and push
       uses: docker/build-push-action@v6
       with:
-        platforms: >-
-          ${{
-              github.event_name == 'push' && github.repository_owner == 'vllm-project' &&
-              'linux/amd64,linux/arm64' ||
-              'linux/arm64'
-          }}
+        platforms: linux/amd64,linux/arm64
         # use the current repo path as the build context, ensure .git is contained
         context: .
         # only trigger when tag, branch/main push

{vllm_ascend-0.9.1rc1 → vllm_ascend-0.9.1rc2}/.github/workflows/image_ubuntu.yml

@@ -1,4 +1,4 @@
-name: 'image / Ubuntu'
+name: 'image'
 # This is a docker build check and publish job:
 # 1. PR Triggered docker image build check
 #   - is for image build check
@@ -39,7 +39,7 @@ on:
 jobs:
 
   build:
-    name: vllm-ascend image build
+    name: vllm-ascend Ubuntu image
     runs-on: ubuntu-latest
 
     steps:
@@ -90,18 +90,12 @@ jobs:
         username: ${{ vars.QUAY_USERNAME }}
         password: ${{ secrets.QUAY_PASSWORD }}
 
-    - name: Build and push 910b
+    - name: Build and push
       uses: docker/build-push-action@v6
       with:
-        platforms: >-
-          ${{
-              github.event_name == 'push' && github.repository_owner == 'vllm-project' &&
-              'linux/amd64,linux/arm64' ||
-              'linux/amd64'
-          }}
+        platforms: linux/amd64,linux/arm64
         # use the current repo path as the build context, ensure .git is contained
         context: .
-        file: Dockerfile
         # only trigger when tag, branch/main push
         push: ${{ github.event_name == 'push' && github.repository_owner == 'vllm-project' }}
         labels: ${{ steps.meta.outputs.labels }}

{vllm_ascend-0.9.1rc1 → vllm_ascend-0.9.1rc2}/.github/workflows/nightly_benchmarks.yaml

@@ -18,11 +18,7 @@
 name: 'Benchmarks / Performance'
 # This workflow runs nightly benchmarks for vllm-ascend.
 
-on:
-  schedule:
-    # Run at 02:00 everyday
-    - cron: '00 18 * * *'
-  
+on:  
   workflow_dispatch:
     # Allow manual triggering of the workflow
 
@@ -51,11 +47,14 @@ jobs:
       matrix:
         include:
           - vllm_branch: v0.9.1
+            vllm_ascend_branch: main
+            vllm_use_v1: 0
+          - vllm_branch: v0.9.0
             vllm_ascend_branch: main
             vllm_use_v1: 1
       max-parallel: 1
     container:
-      image: m.daocloud.io/quay.io/ascend/cann:8.1.rc1-910b-ubuntu22.04-py3.10
+      image: swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/cann:8.2.rc1-910b-ubuntu22.04-py3.11
       volumes:
         - /usr/local/dcmi:/usr/local/dcmi
         - /usr/local/bin/npu-smi:/usr/local/bin/npu-smi
@@ -112,8 +111,6 @@ jobs:
           VLLM_TARGET_DEVICE=empty pip install -e .
 
       - name: Install vllm-project/vllm-ascend
-        env:
-          PIP_EXTRA_INDEX_URL: https://mirrors.huaweicloud.com/ascend/repos/pypi
         run: |
           pip install -e .
           pip install -r benchmarks/requirements-bench.txt
@@ -164,8 +161,6 @@ jobs:
           cp -r benchmarks/* /github/home/benchmarks/
 
       - name: Run benchmark iteration
-        env:
-          PIP_EXTRA_INDEX_URL: https://mirrors.huaweicloud.com/ascend/repos/pypi
         if: github.event_name != 'pull_request'
         run: |
           while IFS= read -r line || [[ -n "$line" ]]; do

{vllm_ascend-0.9.1rc1 → vllm_ascend-0.9.1rc2}/.github/workflows/release_code.yml

@@ -53,7 +53,7 @@ jobs:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        python-version: ["3.10"]
+        python-version: ["3.11"]
     steps:
       - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
 

{vllm_ascend-0.9.1rc1 → vllm_ascend-0.9.1rc2}/.github/workflows/release_whl.yml

@@ -71,16 +71,11 @@ jobs:
         --build-arg PY_VERSION=${{ matrix.python-version }} \
         -t wheel:v1 .
         docker run --rm \
+        -u $(id -u):$(id -g) \
         -v $(pwd):/outpwd \
         wheel:v1 \
         bash -c "cp -r /workspace/vllm-ascend/dist /outpwd"
         ls dist
-      
-    - name: Archive wheel
-      uses: actions/upload-artifact@v4
-      with:
-        name: vllm-ascend-${{ matrix.os }}-py${{ matrix.python-version }}-wheel
-        path: dist/*
 
     - name: Set up Python ${{ matrix.python-version }}
       if: startsWith(github.ref, 'refs/tags/')
@@ -88,6 +83,40 @@ jobs:
       with:
         python-version: ${{ matrix.python-version }}
 
+    - name: Repair wheels with auditwheel
+      run: |
+        python3 -m pip install auditwheel
+        python3 -m pip install patchelf
+        mkdir -p dist/repaired
+        for whl in dist/*.whl; do
+          auditwheel repair "$whl" -w dist/repaired/ \
+          --exclude libplatform.so \
+          --exclude libregister.so \
+          --exclude libge_common_base.so \
+          --exclude libc10.so \
+          --exclude libc_sec.so \
+          --exclude "libascend*.so" \
+          --exclude "libtorch*.so"
+        done
+        rm -f dist/*.whl
+        mv dist/repaired/*.whl dist/
+        rmdir dist/repaired
+        ls dist
+
+    - name: Verify automatic platform tags
+      run: |
+        cd dist
+        for wheel in *.whl; do
+          echo "verification file: $wheel"
+          auditwheel show "$wheel"
+        done
+
+    - name: Archive wheel
+      uses: actions/upload-artifact@v4
+      with:
+        name: vllm-ascend-${{ matrix.os }}-py${{ matrix.python-version }}-wheel
+        path: dist/*
+
     - name: Release
       if: startsWith(github.ref, 'refs/tags/')
       run: |

{vllm_ascend-0.9.1rc1 → vllm_ascend-0.9.1rc2}/.github/workflows/vllm_ascend_doctest.yaml

@@ -29,9 +29,6 @@ on:
       - 'tests/e2e/doctests/**'
       - 'tests/e2e/common.sh'
       - 'tests/e2e/run_doctests.sh'
-  schedule:
-    # Runs every 4 hours
-    - cron:  '0 */4 * * *'
 
 # Bash shells do not use ~/.profile or ~/.bashrc so these shells need to be explicitly
 # declared as "shell: bash -el {0}" on steps that need to be properly activated.
@@ -46,7 +43,7 @@ jobs:
       # Each version should be tested
       fail-fast: false
       matrix:
-        vllm_verison: [v0.9.1-dev, v0.9.1-dev-openeuler, main, main-openeuler]
+        vllm_verison: [main, v0.7.3-dev, main-openeuler, v0.7.3-dev-openeuler]
     name: vLLM Ascend test
     runs-on: linux-arm64-npu-1
     container:
@@ -70,13 +67,13 @@ jobs:
         run: |
           sed -i 's|ports.ubuntu.com|mirrors.tuna.tsinghua.edu.cn|g' /etc/apt/sources.list
           apt-get update -y
-          apt install -y gcc g++ libnuma-dev git curl jq
+          apt install git curl -y
 
       - name: Config OS mirrors - openEuler
         if: ${{ endsWith(matrix.vllm_verison, '-openeuler') }}
         run: |
           yum update -y
-          yum install -y gcc g++ numactl-devel git curl jq
+          yum install git curl -y
 
       - name: Config pip mirrors
         run: |
@@ -88,13 +85,11 @@ jobs:
       - name: Run vllm-ascend/tests/e2e/run_doctests.sh
         run: |
           # PWD: /__w/vllm-ascend/vllm-ascend
-          echo "Replacing /vllm-workspace/vllm-ascend/tests/e2e ..."
-          rm -rf /vllm-workspace/vllm-ascend/tests/e2e
-          mkdir -p /vllm-workspace/vllm-ascend/tests
-          cp -r tests/e2e /vllm-workspace/vllm-ascend/tests/
-
-          # TODO(yikun): Remove this after conf.py merged
-          cp docs/source/conf.py /vllm-workspace/vllm-ascend/docs/source/
+          # Address old branch like v0.7.3:
+          if [ ! -d /vllm-workspace/vllm-ascend/tests/e2e ]; then
+            echo "Warning: the doctest path doesn't exists, copy now"
+            cp -r tests/e2e /vllm-workspace/vllm-ascend/tests/
+          fi
 
           # Simulate container to enter directory
           cd /workspace

vllm_ascend-0.9.1rc2/.github/workflows/vllm_ascend_test.yaml

@@ -0,0 +1,242 @@
+#
+# 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: 'test'
+
+on:
+  pull_request:
+    branches:
+      - 'main'
+      - '*-dev'
+    paths:
+      - '*.txt'
+      - '**/*.py'
+      - '.github/workflows/vllm_ascend_test.yaml'
+      - '!docs/**'
+      - 'pytest.ini'
+      - '!benchmarks/**'
+      - 'tools/mypy.sh'
+      - 'mypy.ini'
+      - '.github/workflows/*.ya?ml'
+      - '.github/workflows/actionlint.*'
+      - '.github/workflows/matchers/actionlint.json'
+
+# Bash shells do not use ~/.profile or ~/.bashrc so these shells need to be explicitly
+# declared as "shell: bash -el {0}" on steps that need to be properly activated.
+# It's used to activate ascend-toolkit environment variables.
+defaults:
+  run:
+    shell: bash -el {0}
+
+jobs:
+  lint:
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        python-version: ["3.11"]
+        vllm_version: [v0.9.1]
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065 # v5.6.0
+        with:
+          python-version: ${{ matrix.python-version }}
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install -r requirements-lint.txt
+      - name: Run codespell check
+        run: |
+          CODESPELL_EXCLUDES=('--skip' 'tests/prompts/**,./benchmarks/sonnet.txt,*tests/lora/data/**,build/**,./vllm_ascend.egg-info/**')
+          CODESPELL_IGNORE_WORDS=('-L' 'CANN,cann,NNAL,nnal,ASCEND,ascend,EnQue,CopyIn')
+
+          codespell --toml pyproject.toml "${CODESPELL_EXCLUDES[@]}" "${CODESPELL_IGNORE_WORDS[@]}"
+      - name: Analysing the code with ruff
+        run: |
+          echo "::add-matcher::.github/workflows/matchers/ruff.json"
+          ruff check --output-format github .
+      - name: Run isort
+        run: |
+          isort . --check-only
+      - name: Running yapf
+        run: |
+          python -m pip install --upgrade pip
+          pip install toml
+          pip install yapf==0.32.0
+          yapf --diff --recursive .
+
+      - name: Install dependencies
+        run: |
+          pip install -r requirements-dev.txt --extra-index-url https://download.pytorch.org/whl/cpu
+
+      - name: Checkout vllm-project/vllm repo
+        uses: actions/checkout@v4
+        with:
+          repository: vllm-project/vllm
+          ref: ${{ matrix.vllm_version }}
+          path: vllm-empty
+
+      - name: Actionlint Check
+        env:
+          SHELLCHECK_OPTS: --exclude=SC2046,SC2006,SC2086
+        run: |
+          echo "::add-matcher::.github/workflows/matchers/actionlint.json"
+          tools/actionlint.sh -color
+
+      - name: Install vllm-project/vllm from source
+        working-directory: vllm-empty
+        run: |
+          pip install -r requirements/build.txt --extra-index-url https://download.pytorch.org/whl/cpu
+          VLLM_TARGET_DEVICE=empty pip install .
+
+      - name: Mypy Check
+        run: |
+          echo "::add-matcher::.github/workflows/matchers/mypy.json"
+          tools/mypy.sh 1 ${{ matrix.python-version }}
+
+  e2e:
+    needs: [lint]
+    if: ${{ needs.lint.result == 'success' }}
+    strategy:
+      max-parallel: 2
+      matrix:
+        os: [linux-arm64-npu-1, linux-arm64-npu-4]
+        vllm_version: [v0.9.1]
+    concurrency:
+      group: >
+        ${{
+        matrix.os == 'linux-arm64-npu-4'
+          && github.event.pull_request.number
+          && format('pr-{0}-limit-npu-4', github.event.pull_request.number)
+        || format('job-{0}-{1}-{2}', matrix.os, matrix.vllm_version, github.event.pull_request.number)
+        }}
+      cancel-in-progress: false
+    name: vLLM Ascend test
+    runs-on: ${{ matrix.os }}
+    container:
+      image: swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/cann:8.2.rc1-910b-ubuntu22.04-py3.11
+      env:
+        HF_ENDPOINT: https://hf-mirror.com
+        HF_TOKEN: ${{ secrets.HF_TOKEN }}
+        VLLM_LOGGING_LEVEL: ERROR
+    steps:
+      - name: Check npu and CANN info
+        run: |
+          npu-smi info
+          cat /usr/local/Ascend/ascend-toolkit/latest/"$(uname -i)"-linux/ascend_toolkit_install.info
+
+      - name: Config mirrors
+        run: |
+          sed -i 's|ports.ubuntu.com|mirrors.tuna.tsinghua.edu.cn|g' /etc/apt/sources.list
+          pip config set global.index-url https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple
+          apt-get update -y
+          apt install git -y
+          git config --global url."https://gh-proxy.test.osinfra.cn/https://github.com/".insteadOf https://github.com/
+
+      - name: Checkout vllm-project/vllm-ascend repo
+        uses: actions/checkout@v4
+
+      - name: Install system dependencies
+        run: |
+          apt-get -y install `cat packages.txt`
+          apt-get -y install gcc g++ cmake libnuma-dev
+
+      - name: Checkout vllm-project/vllm repo
+        uses: actions/checkout@v4
+        with:
+          repository: vllm-project/vllm
+          ref: ${{ matrix.vllm_version }}
+          path: ./vllm-empty
+
+      - name: Install vllm-project/vllm from source
+        working-directory: ./vllm-empty
+        run: |
+          VLLM_TARGET_DEVICE=empty pip install -e .
+
+      - name: Install vllm-project/vllm-ascend
+        run: |
+          pip install -r requirements-dev.txt
+          pip install -v -e .
+
+      - name: Run vllm-project/vllm-ascend test for V1 Engine
+        env:
+          VLLM_USE_V1: 1
+          VLLM_WORKER_MULTIPROC_METHOD: spawn
+        run: |
+          if [[ "${{ matrix.os }}" == "linux-arm64-npu-1" ]]; then
+            VLLM_USE_MODELSCOPE=True pytest -sv tests/singlecard/test_offline_inference.py
+            # guided decoding doesn't work, fix it later
+            # pytest -sv tests/singlecard/test_guided_decoding.py.py
+            # test_ascend_config.py should be ran separately because it will regenerate the global config many times.
+            pytest -sv tests/singlecard/test_ascend_config.py
+            pytest -sv tests/singlecard/test_camem.py
+            pytest -sv tests/singlecard/core/test_ascend_scheduler.py
+            pytest -sv tests/singlecard/core/test_ascend_scheduler_e2e.py
+            pytest -sv tests/singlecard/ \
+            --ignore=tests/singlecard/test_offline_inference.py \
+            --ignore=tests/singlecard/test_guided_decoding.py \
+            --ignore=tests/singlecard/test_ascend_config.py \
+            --ignore=tests/singlecard/test_camem.py \
+            --ignore=tests/singlecard/core/test_ascend_scheduler.py \
+            --ignore=tests/singlecard/core/test_ascend_scheduler_e2e.py
+          else
+            pytest -sv tests/multicard/test_ilama_lora_tp2.py
+            # To avoid oom, we need to run the test in a single process.
+            VLLM_USE_MODELSCOPE=True pytest -sv tests/multicard/test_w4a8_deepseek.py::test_deepseek_W4A8
+            VLLM_USE_MODELSCOPE=True pytest -sv tests/multicard/test_offline_inference_distributed.py::test_models_distributed_QwQ
+            VLLM_USE_MODELSCOPE=True pytest -sv tests/multicard/test_offline_inference_distributed.py::test_models_distributed_DeepSeek
+            VLLM_USE_MODELSCOPE=True pytest -sv tests/multicard/test_offline_inference_distributed.py::test_models_distributed_topk
+            VLLM_USE_MODELSCOPE=True pytest -sv tests/multicard/test_offline_inference_distributed.py::test_models_distributed_DeepSeek_W8A8
+            VLLM_USE_MODELSCOPE=True pytest -sv tests/multicard/test_offline_inference_distributed.py::test_models_distributed_DeepSeek_dbo
+            VLLM_USE_MODELSCOPE=True pytest -sv tests/multicard/test_offline_inference_distributed.py::test_models_distributed_DeepSeek_w8a8_ep_dbo
+            VLLM_USE_MODELSCOPE=True pytest -sv tests/multicard/test_offline_inference_distributed.py::test_models_distributed_DeepSeekV3_dbo
+            VLLM_USE_MODELSCOPE=True pytest -sv tests/multicard/test_offline_inference_distributed.py::test_models_distributed_QwQ_with_flashcomm_v1
+            VLLM_USE_MODELSCOPE=True pytest -sv tests/multicard/test_offline_inference_distributed.py::test_models_distributed_Qwen3_with_flashcomm_v2
+            VLLM_USE_MODELSCOPE=True pytest -sv tests/multicard/ --ignore=tests/multicard/test_ilama_lora_tp2.py --ignore=tests/multicard/test_offline_inference_distributed.py --ignore=tests/multicard/test_w4a8_deepseek.py
+          fi
+
+      - name: Run vllm-project/vllm-ascend test on V0 engine
+        if: ${{ github.event_name == 'schedule' }}
+        env:
+          VLLM_USE_V1: 0
+        run: |
+          if [[ "${{ matrix.os }}" == "linux-arm64-npu-1" ]]; then
+            VLLM_USE_MODELSCOPE=True  pytest -sv tests/singlecard/test_offline_inference.py
+            # guided decoding doesn't work, fix it later
+            # pytest -sv tests/singlecard/test_guided_decoding.py.py
+            pytest -sv tests/singlecard/test_camem.py
+            # test_ascend_config.py should be ran separately because it will regenerate the global config many times.
+            pytest -sv tests/singlecard/test_ascend_config.py
+            pytest -sv tests/singlecard/test_prompt_embedding.py
+            pytest -sv tests/singlecard/ \
+              --ignore=tests/singlecard/test_offline_inference.py \
+              --ignore=tests/singlecard/test_guided_decoding.py \
+              --ignore=tests/singlecard/test_camem.py \
+              --ignore=tests/singlecard/test_ascend_config.py \
+              --ignore=tests/singlecard/test_prompt_embedding.py \
+              --ignore=tests/singlecard/core/test_ascend_scheduler.py \
+              --ignore=tests/singlecard/core/test_ascend_scheduler_e2e.py
+          else
+            pytest -sv tests/multicard/test_ilama_lora_tp2.py
+            # Fixme: run VLLM_USE_MODELSCOPE=True pytest -sv tests/multicard/test_offline_inference_distributed.py will raise error.
+            # To avoid oom, we need to run the test in a single process.
+            VLLM_USE_MODELSCOPE=True pytest -sv tests/multicard/test_offline_inference_distributed.py::test_models_distributed_QwQ
+            VLLM_USE_MODELSCOPE=True pytest -sv tests/multicard/test_offline_inference_distributed.py::test_models_distributed_DeepSeek
+            VLLM_USE_MODELSCOPE=True pytest -sv tests/multicard/test_offline_inference_distributed.py::test_models_distributed_topk
+            VLLM_USE_MODELSCOPE=True pytest -sv tests/multicard/test_offline_inference_distributed.py::test_models_distributed_DeepSeek_W8A8
+            VLLM_USE_MODELSCOPE=True pytest -sv tests/multicard/ --ignore=tests/multicard/test_ilama_lora_tp2.py --ignore=tests/multicard/test_offline_inference_distributed.py
+          fi

{vllm_ascend-0.9.1rc1 → vllm_ascend-0.9.1rc2}/.github/workflows/vllm_ascend_test_long_term.yaml

@@ -17,9 +17,6 @@
 name: 'e2e test / long-term-test'
 
 on:
-  schedule:
-    # Runs at 23:00 UTC (7:00 AM Beijing) every day
-    - cron: '0 23 * * *'
   pull_request:
     types: [ labeled ]
 
@@ -43,12 +40,12 @@ jobs:
       max-parallel: 2
       matrix:
         os: [linux-arm64-npu-1, linux-arm64-npu-4]
-        vllm_version: [main, v0.9.1]
+        vllm_version: [v0.9.1]
     name: vLLM Ascend long term test
     runs-on: ${{ matrix.os }}
     container:
       # TODO(yikun): Remove m.daocloud.io prefix when infra proxy ready
-      image: m.daocloud.io/quay.io/ascend/cann:8.1.rc1-910b-ubuntu22.04-py3.10
+      image: swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/cann:8.2.rc1-910b-ubuntu22.04-py3.11
       env:
         HF_ENDPOINT: https://hf-mirror.com
         HF_TOKEN: ${{ secrets.HF_TOKEN }}
@@ -88,8 +85,6 @@ jobs:
           VLLM_TARGET_DEVICE=empty pip install -e .
 
       - name: Install vllm-project/vllm-ascend
-        env:
-          PIP_EXTRA_INDEX_URL: https://mirrors.huaweicloud.com/ascend/repos/pypi
         run: |
           pip install -r requirements-dev.txt
           pip install -v -e .
@@ -97,13 +92,17 @@ jobs:
       - name: Run vllm-project/vllm-ascend long term test
         run: |
           if [[ "${{ matrix.os }}" == "linux-arm64-npu-1" ]]; then
-            # spec decode test
-            VLLM_USE_MODELSCOPE=True pytest -sv tests/e2e/long_term/spec_decode/e2e/test_v1_mtp_correctness.py
+            # v0 spec decode test
+            # VLLM_USE_MODELSCOPE=True pytest -sv tests/long_term/spec_decode_v0/e2e/test_mtp_correctness.py  # it needs a clean process
+            # pytest -sv tests/long_term/spec_decode_v0 --ignore=tests/long_term/spec_decode_v0/e2e/test_mtp_correctness.py
+            # v1 spec decode test
+            # TODO: revert me when test_v1_mtp_correctness.py is fixed
+            VLLM_USE_MODELSCOPE=True pytest -sv tests/long_term/spec_decode_v1/test_v1_mtp_correctness.py
             # TODO: revert me when test_v1_spec_decode.py::test_ngram_correctness is fixed
-            VLLM_USE_MODELSCOPE=True pytest -sv tests/e2e/long_term/spec_decode/e2e/test_v1_spec_decode.py
-            VLLM_USE_MODELSCOPE=True pytest -sv tests/e2e/long_term/spec_decode/e2e/test_mtp_correctness.py  # it needs a clean process
-            pytest -sv tests/e2e/long_term/spec_decode --ignore=tests/e2e/long_term/spec_decode/e2e/test_mtp_correctness.py --ignore=tests/e2e/long_term/spec_decode/e2e/test_v1_spec_decode.py --ignore=tests/e2e/long_term/spec_decode/e2e/test_v1_mtp_correctness.py
-            pytest -sv tests/e2e/long_term/test_accuracy.py
+            # VLLM_USE_MODELSCOPE=True pytest -sv tests/long_term/spec_decode_v1/test_v1_spec_decode.py
+            # accuracy test single card
+            pytest -sv tests/long_term/test_accuracy.py
           else
-            VLLM_USE_MODELSCOPE=True pytest -sv tests/e2e/long_term/test_deepseek_v2_lite_tp2_accuracy.py
+            # accuracy test multi card
+            VLLM_USE_MODELSCOPE=True pytest -sv tests/long_term/test_deepseek_v2_lite_tp2_accuracy.py
           fi

{vllm_ascend-0.9.1rc1 → vllm_ascend-0.9.1rc2}/.github/workflows/vllm_ascend_test_pd.yaml

@@ -17,9 +17,6 @@
 name: 'e2e test / pd-disaggregation'
 
 on:
-  schedule:
-    # Runs at 23:00 UTC (7:00 AM Beijing) every day
-    - cron: '0 23 * * *'
   pull_request:
     types: [ labeled ]
 
@@ -41,12 +38,12 @@ jobs:
     if: ${{ contains(github.event.pull_request.labels.*.name, 'pd-test') && contains(github.event.pull_request.labels.*.name, 'ready-for-test') || github.event_name == 'schedule' }}
     strategy:
       matrix:
-        vllm_verison: [main, v0.9.1]
+        vllm_verison: [v0.9.1]
     name: vLLM Ascend prefilling decoding disaggregation test
     runs-on: linux-arm64-npu-static-8
 
     container:
-      image: m.daocloud.io/quay.io/ascend/cann:8.1.rc1-910b-ubuntu22.04-py3.10
+      image: swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/cann:8.2.rc1-910b-ubuntu22.04-py3.11
       volumes:
         - /usr/local/dcmi:/usr/local/dcmi
         - /usr/local/bin/npu-smi:/usr/local/bin/npu-smi
@@ -97,8 +94,6 @@ jobs:
           VLLM_TARGET_DEVICE=empty pip install -e .
 
       - name: Install vllm-project/vllm-ascend
-        env:
-          PIP_EXTRA_INDEX_URL: https://mirrors.huaweicloud.com/ascend/repos/pypi
         run: |
           pip install -r requirements-dev.txt
           pip install -v -e .
@@ -106,3 +101,7 @@ jobs:
       - name: Run vllm-project/vllm-ascend PD Disaggregation test
         run: |
           pytest -sv tests/e2e/pd_disaggreate/test_pd_e2e.py
+      
+      - name: Run vllm-project/vllm-ascend PD Disaggregation edge test
+        run: |
+          bash tests/e2e/pd_disaggreate/run_edge_case_test.sh