vllm-ascend 0.9.0rc2__tar.gz → 0.9.1rc2__tar.gz

{vllm_ascend-0.9.0rc2 → 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-0.9.0rc2 → 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-0.9.0rc2 → vllm_ascend-0.9.1rc2}/.github/workflows/accuracy_test.yaml

@@ -34,8 +34,7 @@ on:
         # Current supported vLLM versions
         options:
           - main
-          - v0.9.0.1
-          - v0.9.0
+          - v0.9.1
           - v0.7.3
       vllm-ascend-version:
         description: 'vllm-ascend version:'
@@ -118,7 +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 }}
@@ -159,7 +158,7 @@ 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 || 'v0.9.1' }}
 
       - name: Install vllm-project/vllm from source
         working-directory: ./vllm-empty

{vllm_ascend-0.9.0rc2 → vllm_ascend-0.9.1rc2}/.github/workflows/image_openeuler.yml

@@ -19,6 +19,12 @@ on:
       - '.github/workflows/image_openeuler.yml'
       - 'Dockerfile.openEuler'
       - 'vllm_ascend/**'
+      - 'setup.py'
+      - 'pyproject.toml'
+      - 'requirements.txt'
+      - 'cmake/**'
+      - 'CMakeLists.txt'
+      - 'csrc/**'
   push:
     # Publish image when tagging, the Dockerfile in tag will be build as tag image
     branches:

{vllm_ascend-0.9.0rc2 → vllm_ascend-0.9.1rc2}/.github/workflows/image_ubuntu.yml

@@ -19,6 +19,12 @@ on:
       - '.github/workflows/image_ubuntu.yml'
       - 'Dockerfile'
       - 'vllm_ascend/**'
+      - 'setup.py'
+      - 'pyproject.toml'
+      - 'requirements.txt'
+      - 'cmake/**'
+      - 'CMakeLists.txt'
+      - 'csrc/**'
   push:
     # Publish image when tagging, the Dockerfile in tag will be build as tag image
     branches:

{vllm_ascend-0.9.0rc2 → 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
 
@@ -45,15 +41,20 @@ jobs:
   test:
     if: ${{ contains(github.event.pull_request.labels.*.name, 'performance-test') && contains(github.event.pull_request.labels.*.name, 'ready-for-test') || github.event_name == 'schedule' || github.event_name == 'workflow_dispatch' }}
 
-    name: Benchmarks/vLLM=${{ matrix.vllm_branch }}, vLLM-Ascend=${{ matrix.vllm_ascend_branch }}
+    name: Benchmarks/vLLM=${{ matrix.vllm_branch }}, vLLM-Ascend=${{ matrix.vllm_ascend_branch }}, use_v1=${{ matrix.vllm_use_v1 }}
     runs-on: 'linux-arm64-npu-static-8'
     strategy:
       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
@@ -71,6 +72,7 @@ jobs:
         HF_TOKEN: ${{ secrets.HF_TOKEN }}
         ES_OM_DOMAIN: ${{ secrets.ES_OM_DOMAIN }}
         ES_OM_AUTHORIZATION: ${{ secrets.ES_OM_AUTHORIZATION }}
+        VLLM_USE_V1: ${{ matrix.vllm_use_v1 }}
     steps:
       - name: Check npu and CANN info
         run: |
@@ -140,7 +142,7 @@ jobs:
       - name: Install elastic_tool
         if: github.event_name != 'pull_request'
         run: |
-          pip install escli-tool==0.2.1
+          pip install escli-tool==0.2.2
           
       - name: Collect pr info from vllm-project/vllm-ascend
         if: github.event_name != 'pull_request'
@@ -164,10 +166,10 @@ jobs:
           while IFS= read -r line || [[ -n "$line" ]]; do
             commit_id=${line%% *}
             commit_title=${line#* }
-            commit_time=$(git show -s --format=%cd $commit_hash --date=iso-strict)
-            commit_time_no_tz=${commit_time::19}
 
             git checkout $commit_id
+            commit_time=$(git show -s --format=%cd $commit_hash --date=iso-strict)
+            commit_time_no_tz=${commit_time::19}
             pip install -e .
             
             echo "------------------------"
@@ -177,17 +179,17 @@ jobs:
             echo "vllm branch: ${{ matrix.vllm_branch }}"
             echo "vllm-ascend branch: ${{ matrix.vllm_ascend_branch }}"
             echo "------------------------"
+            
             cd /github/home
             bash benchmarks/scripts/run-performance-benchmarks.sh
             # send the result to es
-            if [[ "${{ github.event_name }}" != "pull request" ]]; then
-              escli add --vllm_branch ${{ matrix.vllm_branch }} \
-              --vllm_ascend_branch ${{ matrix.vllm_ascend_branch }} \
-              --commit_id $commit_id \
-              --commit_title "$commit_title" \
-              --created_at "$commit_time_no_tz" \
-              --res_dir ./benchmarks/results 
-              rm -rf ./benchmarks/results
-            fi
+            escli add --vllm_branch ${{ matrix.vllm_branch }} \
+            --vllm_ascend_branch ${{ matrix.vllm_ascend_branch }} \
+            --commit_id $commit_id \
+            --commit_title "$commit_title" \
+            --created_at "$commit_time_no_tz" \
+            --res_dir ./benchmarks/results \
+            --extra_feat '{"VLLM_USE_V1": "${{ matrix.vllm_use_v1 }}"}'
+            rm -rf ./benchmarks/results
             cd -
           done < commit_log.txt

{vllm_ascend-0.9.0rc2 → 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-0.9.0rc2 → 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-0.9.0rc2 → 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.

{vllm_ascend-0.9.0rc2 → vllm_ascend-0.9.1rc2}/.github/workflows/vllm_ascend_test.yaml

@@ -18,8 +18,6 @@
 name: 'test'
 
 on:
-  schedule:
-    - cron: '0 23 * * *'
   pull_request:
     branches:
       - 'main'
@@ -33,6 +31,9 @@ on:
       - '!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.
@@ -46,7 +47,8 @@ jobs:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        python-version: ["3.10"]
+        python-version: ["3.11"]
+        vllm_version: [v0.9.1]
     steps:
       - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
       - name: Set up Python ${{ matrix.python-version }}
@@ -85,8 +87,16 @@ jobs:
         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: |
@@ -105,7 +115,7 @@ jobs:
       max-parallel: 2
       matrix:
         os: [linux-arm64-npu-1, linux-arm64-npu-4]
-        vllm_version: [main, v0.9.0]
+        vllm_version: [v0.9.1]
     concurrency:
       group: >
         ${{
@@ -118,8 +128,7 @@ jobs:
     name: vLLM Ascend 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: ${{ secrets.HF_TOKEN }}
@@ -170,34 +179,43 @@ jobs:
         run: |
           if [[ "${{ matrix.os }}" == "linux-arm64-npu-1" ]]; then
             VLLM_USE_MODELSCOPE=True pytest -sv tests/singlecard/test_offline_inference.py
-            pytest -sv tests/singlecard/test_scheduler.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_scheduler.py \
             --ignore=tests/singlecard/test_guided_decoding.py \
             --ignore=tests/singlecard/test_ascend_config.py \
-            --ignore=tests/singlecard/test_camem.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/ --ignore=tests/multicard/test_ilama_lora_tp2.py --ignore=tests/multicard/test_offline_inference_distributed.py
+            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
-            pytest -sv tests/singlecard/test_scheduler.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
@@ -206,11 +224,12 @@ jobs:
             pytest -sv tests/singlecard/test_prompt_embedding.py
             pytest -sv tests/singlecard/ \
               --ignore=tests/singlecard/test_offline_inference.py \
-              --ignore=tests/singlecard/test_scheduler.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/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.
@@ -218,5 +237,6 @@ jobs:
             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-0.9.0rc2 → 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, v0.9.0]
+        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: ${{ secrets.HF_TOKEN }}
@@ -95,12 +92,17 @@ jobs:
       - name: Run vllm-project/vllm-ascend long term test
         run: |
           if [[ "${{ matrix.os }}" == "linux-arm64-npu-1" ]]; then
-            # spec decode test
-            VLLM_USE_MODELSCOPE=True pytest -sv tests/long_term/spec_decode/e2e/test_v1_mtp_correctness.py
-            VLLM_USE_MODELSCOPE=True pytest -sv tests/long_term/spec_decode/e2e/test_v1_spec_decode.py
-            VLLM_USE_MODELSCOPE=True pytest -sv tests/long_term/spec_decode/e2e/test_mtp_correctness.py  # it needs a clean process
-            pytest -sv tests/long_term/spec_decode --ignore=tests/long_term/spec_decode/e2e/test_mtp_correctness.py --ignore=tests/long_term/spec_decode/e2e/test_v1_spec_decode.py --ignore=tests/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/long_term/spec_decode_v1/test_v1_spec_decode.py
+            # accuracy test single card
             pytest -sv tests/long_term/test_accuracy.py
           else
+            # accuracy test multi card
             VLLM_USE_MODELSCOPE=True pytest -sv tests/long_term/test_deepseek_v2_lite_tp2_accuracy.py
           fi

{vllm_ascend-0.9.0rc2 → 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, v0.9.0]
+        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
@@ -104,3 +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

{vllm_ascend-0.9.0rc2 → vllm_ascend-0.9.1rc2}/CMakeLists.txt

@@ -96,5 +96,3 @@ target_link_libraries(
 target_link_options(vllm_ascend_C PRIVATE "-Wl,-rpath,$ORIGIN:$ORIGIN/lib")
 
 install(TARGETS vllm_ascend_C vllm_ascend_kernels DESTINATION ${VLLM_ASCEND_INSTALL_PATH})
-
-

{vllm_ascend-0.9.0rc2 → vllm_ascend-0.9.1rc2}/Dockerfile

@@ -15,7 +15,7 @@
 # This file is a part of the vllm-ascend project.
 #
 
-FROM quay.io/ascend/cann:8.1.rc1-910b-ubuntu22.04-py3.10
+FROM quay.io/ascend/cann:8.2.rc1-910b-ubuntu22.04-py3.11
 
 ARG PIP_INDEX_URL="https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple"
 ARG COMPILE_CUSTOM_KERNELS=1
@@ -37,7 +37,7 @@ RUN pip config set global.index-url ${PIP_INDEX_URL}
 
 # Install vLLM
 ARG VLLM_REPO=https://github.com/vllm-project/vllm.git
-ARG VLLM_TAG=v0.9.0
+ARG VLLM_TAG=v0.9.1
 RUN git clone --depth 1 $VLLM_REPO --branch $VLLM_TAG /vllm-workspace/vllm
 # In x86, triton will be installed by vllm. But in Ascend, triton doesn't work correctly. we need to uninstall it.
 RUN VLLM_TARGET_DEVICE="empty" python3 -m pip install -v -e /vllm-workspace/vllm/ --extra-index https://download.pytorch.org/whl/cpu/ && \

{vllm_ascend-0.9.0rc2 → vllm_ascend-0.9.1rc2}/Dockerfile.openEuler

@@ -15,7 +15,7 @@
 # This file is a part of the vllm-ascend project.
 #
 
-FROM quay.io/ascend/cann:8.1.rc1-910b-openeuler22.03-py3.10
+FROM quay.io/ascend/cann:8.2.rc1-910b-openeuler22.03-py3.11
 
 ARG PIP_INDEX_URL="https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple"
 ARG COMPILE_CUSTOM_KERNELS=1
@@ -34,7 +34,7 @@ COPY . /vllm-workspace/vllm-ascend/
 
 # Install vLLM
 ARG VLLM_REPO=https://github.com/vllm-project/vllm.git
-ARG VLLM_TAG=v0.9.0
+ARG VLLM_TAG=v0.9.1
 
 RUN git clone --depth 1 $VLLM_REPO --branch $VLLM_TAG /vllm-workspace/vllm
 # In x86, triton will be installed by vllm. But in Ascend, triton doesn't work correctly. we need to uninstall it.

{vllm_ascend-0.9.0rc2 → vllm_ascend-0.9.1rc2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: vllm_ascend
-Version: 0.9.0rc2
+Version: 0.9.1rc2
 Summary: vLLM Ascend backend plugin
 Home-page: https://github.com/vllm-project/vllm-ascend
 Author: vLLM-Ascend team
@@ -58,8 +58,8 @@ By using vLLM Ascend plugin, popular open-source models, including Transformer-l
 - OS: Linux
 - Software:
   * Python >= 3.9, < 3.12
-  * CANN >= 8.1.RC1
-  * PyTorch >= 2.5.1, torch-npu >= 2.5.1
+  * CANN >= 8.2.RC1
+  * PyTorch >= 2.5.1, torch-npu >= 2.5.1.post1
   * vLLM (the same version as vllm-ascend)
 
 ## Getting Started

{vllm_ascend-0.9.0rc2 → vllm_ascend-0.9.1rc2}/README.md

@@ -37,8 +37,8 @@ By using vLLM Ascend plugin, popular open-source models, including Transformer-l
 - OS: Linux
 - Software:
   * Python >= 3.9, < 3.12
-  * CANN >= 8.1.RC1
-  * PyTorch >= 2.5.1, torch-npu >= 2.5.1
+  * CANN >= 8.2.RC1
+  * PyTorch >= 2.5.1, torch-npu >= 2.5.1.post1
   * vLLM (the same version as vllm-ascend)
 
 ## Getting Started

{vllm_ascend-0.9.0rc2 → vllm_ascend-0.9.1rc2}/README.zh.md

@@ -38,8 +38,8 @@ vLLM 昇腾插件 (`vllm-ascend`) 是一个由社区维护的让vLLM在Ascend NP
 - 操作系统：Linux
 - 软件：
   * Python >= 3.9, < 3.12
-  * CANN >= 8.1.RC1
-  * PyTorch >= 2.5.1, torch-npu >= 2.5.1
+  * CANN >= 8.2.RC1
+  * PyTorch >= 2.5.1, torch-npu >= 2.5.1.post1
   * vLLM (与vllm-ascend版本一致)
 
 ## 开始使用