vllm-ascend 0.13.0__tar.gz → 0.13.0rc2__tar.gz

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/.github/Dockerfile.buildwheel

@@ -32,7 +32,7 @@ COPY . /workspace/vllm-ascend/
 
 # Install req
 RUN python3 -m pip install -r vllm-ascend/requirements.txt --extra-index https://download.pytorch.org/whl/cpu/ && \
-    python3 -m pip install twine attrs psutil
+    python3 -m pip install twine
 
 # Install vllm-ascend
 RUN source /usr/local/Ascend/ascend-toolkit/set_env.sh && \

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/.github/Dockerfile.nightly.a2

@@ -15,7 +15,7 @@
 # This file is a part of the vllm-ascend project.
 #
 
-FROM quay.io/ascend/vllm-ascend:releases-v0.13.0
+FROM quay.io/ascend/vllm-ascend:main
 
 ARG PIP_INDEX_URL="https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple"
 ARG AIS_BENCH_TAG="v3.0-20250930-master"

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/.github/Dockerfile.nightly.a3

@@ -15,7 +15,7 @@
 # This file is a part of the vllm-ascend project.
 #
 
-FROM quay.io/ascend/vllm-ascend:releases-v0.13.0-a3
+FROM quay.io/ascend/vllm-ascend:main-a3
 
 ARG PIP_INDEX_URL="https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple"
 ARG AIS_BENCH_TAG="v3.0-20250930-master"

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/.github/actionlint.yaml

@@ -21,7 +21,3 @@ self-hosted-runner:
     - linux-aarch64-a3-0
     - linux-amd64-cpu-8-hk
     - linux-amd64-cpu-16-hk
-    - linux-aarch64-a2b3-0
-    - linux-aarch64-a2b3-1
-    - linux-aarch64-a2b3-2
-    - linux-aarch64-a2b3-4

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/.github/workflows/_e2e_nightly_single_node.yaml

@@ -110,6 +110,19 @@ jobs:
           fi
           cd ..
 
+      - name: Install custom-ops (for DeepSeek-V3.2-Exp)
+        if: ${{ inputs.name == 'deepseek3_2-exp-w8a8' }}
+        shell: bash -l {0}
+        run: |
+          wget https://vllm-ascend.obs.cn-north-4.myhuaweicloud.com/vllm-ascend/a3/CANN-custom_ops-sfa-linux.aarch64.run
+          chmod +x ./CANN-custom_ops-sfa-linux.aarch64.run
+          ./CANN-custom_ops-sfa-linux.aarch64.run --quiet
+          export ASCEND_CUSTOM_OPP_PATH=/usr/local/Ascend/ascend-toolkit/latest/opp/vendors/customize:${ASCEND_CUSTOM_OPP_PATH}
+          export LD_LIBRARY_PATH=/usr/local/Ascend/ascend-toolkit/latest/opp/vendors/customize/op_api/lib/:${LD_LIBRARY_PATH}
+          wget https://vllm-ascend.obs.cn-north-4.myhuaweicloud.com/vllm-ascend/a3/custom_ops-1.0-cp311-cp311-linux_aarch64.whl
+          pip install custom_ops-1.0-cp311-cp311-linux_aarch64.whl
+          . /usr/local/Ascend/ascend-toolkit/set_env.sh
+
       - name: Run vllm-project/vllm-ascend test
         env:
           VLLM_WORKER_MULTIPROC_METHOD: spawn
@@ -121,3 +134,5 @@ jobs:
           # ignore test_dispatch_ffn_combine until the test is fixed
           pytest -sv ${{ inputs.tests }} \
           --ignore=tests/e2e/nightly/single_node/ops/singlecard_ops/test_fused_moe.py
+
+

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/.github/workflows/_e2e_nightly_single_node_models.yaml

@@ -112,10 +112,10 @@ jobs:
           update-alternatives --install /usr/bin/clang++ clang++ /usr/bin/clang++-15 20
 
       - name: Install tensorflow (for Molmo-7B-D-0924)
-        if: ${{ inputs.runner == 'linux-aarch64-a2b3-1' && contains(inputs.model_list, 'Molmo-7B-D-0924') }}
+        if: ${{ inputs.runner == 'linux-aarch64-a2-1' && contains(inputs.model_list, 'Molmo-7B-D-0924') }}
         shell: bash -l {0}
         run: |
-          pip install tensorflow==2.19.1 --no-cache-dir
+          pip install tensorflow --no-cache-dir
 
       - name: Resolve vllm-ascend version
         run: |

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/.github/workflows/_e2e_test.yaml

@@ -19,7 +19,7 @@ on:
 jobs:
   e2e:
     name: singlecard
-    runs-on: linux-aarch64-a2b3-1
+    runs-on: ${{ inputs.runner }}-1
     container:
       image: ${{ inputs.image }}
       env:
@@ -145,9 +145,11 @@ jobs:
 
       - name: Config mirrors
         run: |
-          sed -Ei 's@(ports|archive).ubuntu.com@cache-service.nginx-pypi-cache.svc.cluster.local:8081@g' /etc/apt/sources.list
-          pip config set global.index-url http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple
-          pip config set global.trusted-host cache-service.nginx-pypi-cache.svc.cluster.local
+          # Fix me: use nginx cache rather than the pypi
+          # sed -Ei 's@(ports|archive).ubuntu.com@cache-service.nginx-pypi-cache.svc.cluster.local:8081@g' /etc/apt/sources.list
+          # pip config set global.index-url http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple
+          # pip config set global.trusted-host cache-service.nginx-pypi-cache.svc.cluster.local
+          pip config set global.index-url https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple
           apt-get update -y
           apt install git -y
 

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/.github/workflows/_schedule_image_build.yaml

@@ -46,7 +46,6 @@ jobs:
       with:
         fetch-depth: 0
         persist-credentials: false
-        ref: ${{ github.ref }}
 
     - name: Free up disk space
       uses: jlumbroso/free-disk-space@54081f138730dfa15788a46383842cd2f914a1be # v1.3.1
@@ -106,8 +105,6 @@ jobs:
     steps:
       - name: Checkout
         uses: actions/checkout@v6
-        with:
-          ref: ${{ github.ref }}
 
       - name: Download arm64 digests
         uses: actions/download-artifact@v7
@@ -149,9 +146,8 @@ jobs:
           #      which follow the rule from vLLM with prefix v
           # TODO(yikun): the post release might be considered as latest release
           tags: |
-            type=ref,event=branch,suffix=${{ env.SUFFIX }}
-            type=ref,event=pr,suffix=${{ env.SUFFIX }}
             type=pep440,pattern={{raw}},suffix=${{ env.SUFFIX }}
+            type=schedule,pattern=main,suffix=${{ env.SUFFIX }}
           flavor:
             latest=false
 

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/.github/workflows/_unit_test.yaml

@@ -27,7 +27,6 @@ jobs:
         VLLM_USE_MODELSCOPE: True
         SOC_VERSION: ascend910b1
         MAX_JOBS: 4
-        COMPILE_CUSTOM_KERNELS: 0
     steps:
       - name: Install packages
         run: |

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/.github/workflows/labled_doctest.yaml

@@ -48,7 +48,7 @@ jobs:
       matrix:
         vllm_verison: [v0.9.1-dev, v0.9.1-dev-openeuler, main, main-openeuler]
     name: vLLM Ascend test
-    runs-on: linux-aarch64-a2b3-1
+    runs-on: linux-aarch64-a2-1
     container:
       image: swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/vllm-ascend:${{ matrix.vllm_verison }}
     steps:

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/.github/workflows/labled_download_model.yaml

@@ -19,7 +19,7 @@ jobs:
   download-models:
     if: contains(github.event.pull_request.labels.*.name, 'model-download')
     name: Download models from ModelScope
-    runs-on: linux-aarch64-a2b3-0
+    runs-on: linux-aarch64-a2-0
     container:
       image: swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/vllm-ascend:nightly-cpu
 

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/.github/workflows/nightly_test_a2.yaml

@@ -50,22 +50,22 @@ jobs:
       matrix:
         test_config:
           - name: qwen3-8b
-            os: linux-aarch64-a2b3-1
+            os: linux-aarch64-a2-1
             tests: tests/e2e/nightly/single_node/models/test_qwen3_8b.py
           - name: qwen3next
-            os: linux-aarch64-a2b3-4
+            os: linux-aarch64-a2-4
             ests: tests/e2e/nightly/single_node/models/test_qwen3_next.py
           - name: qwen3-32b
-            os: linux-aarch64-a2b3-4
+            os: linux-aarch64-a2-4
             tests: tests/e2e/nightly/single_node/models/test_qwen3_32b.py
           - name: qwen3-32b-in8-a2
-            os: linux-aarch64-a2b3-4
+            os: linux-aarch64-a2-4
             tests: tests/e2e/nightly/single_node/models/test_qwen3_32b_int8.py
           - name: test_custom_op
-            os: linux-aarch64-a2b3-1
+            os: linux-aarch64-a2-1
             tests: tests/e2e/nightly/single_node/ops/singlecard_ops
           - name: test_custom_op_multi_card
-            os: linux-aarch64-a2b3-4
+            os: linux-aarch64-a2-4
             tests: tests/e2e/nightly/single_node/ops/multicard_ops_a2/
     uses: ./.github/workflows/_e2e_nightly_single_node.yaml
     with:
@@ -93,7 +93,7 @@ jobs:
     uses: ./.github/workflows/_e2e_nightly_multi_node.yaml
     with:
       soc_version: a2
-      runner: linux-aarch64-a2b3-0
+      runner: linux-aarch64-a2-0
       image: 'swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/vllm-ascend:nightly-a2'
       replicas: 1
       size: ${{ matrix.test_config.size }}
@@ -106,32 +106,35 @@ jobs:
       ${{
         github.event_name == 'schedule' ||
         github.event_name == 'workflow_dispatch' ||
-        contains(github.event.pull_request.labels.*.name, 'accuracy-test')
+        (
+          contains(github.event.pull_request.labels.*.name, 'accuracy-test') &&
+          contains(github.event.pull_request.labels.*.name, 'ready-for-test')
+        )
       }}
     strategy:
       fail-fast: false
       matrix:
         test_config:
-          - os: linux-aarch64-a2b3-1
+          - os: linux-aarch64-a2-1
             model_list:
               - Qwen3-8B
               - Qwen2-Audio-7B-Instruct
               - Qwen3-8B-W8A8
               - Qwen3-VL-8B-Instruct
               - Qwen2.5-Omni-7B
-          - os: linux-aarch64-a2b3-1
+          - os: linux-aarch64-a2-1
             model_list:
               - ERNIE-4.5-21B-A3B-PT
               - InternVL3_5-8B-hf
               - Molmo-7B-D-0924
               - Llama-3.2-3B-Instruct
               - llava-onevision-qwen2-0.5b-ov-hf
-          - os: linux-aarch64-a2b3-2
+          - os: linux-aarch64-a2-2
             model_list:
               - Qwen3-30B-A3B
               - Qwen3-VL-30B-A3B-Instruct
               - Qwen3-30B-A3B-W8A8
-          - os: linux-aarch64-a2b3-4
+          - os: linux-aarch64-a2-4
             model_list:
               - Qwen3-Next-80B-A3B-Instruct
               - Qwen3-Omni-30B-A3B-Instruct

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/.github/workflows/nightly_test_a3.yaml

@@ -83,12 +83,6 @@ jobs:
           - name: multi-node-qwen-vl-disagg-pd
             config_file_path: Qwen3-VL-235B-disagg-pd.yaml
             size: 2
-          - name: multi-node-kimi-k2-instruct-w8a8
-            config_file_path: Kimi-K2-Instruct-W8A8.yaml
-            size: 2
-          - name: multi-node-deepseek-v3.1
-            config_file_path: DeepSeek-V3.1-BF16.yaml
-            size: 2
     uses: ./.github/workflows/_e2e_nightly_multi_node.yaml
     with:
       soc_version: a3
@@ -150,15 +144,10 @@ jobs:
           - name: qwen3-next-w8a8
             os: linux-aarch64-a3-4
             tests: tests/e2e/nightly/single_node/models/test_qwen3_next_w8a8.py
-          - name: kimi-k2-thinking
-            os: linux-aarch64-a3-16
-            tests: tests/e2e/nightly/single_node/models/test_kimi_k2_thinking.py
-          - name: deepseek-r1-w8a8-hbm
-            os: linux-aarch64-a3-16
-            tests: tests/e2e/nightly/single_node/models/test_deepseek_r1_w8a8_hbm.py
-          - name: deepseek3_2-w8a8
-            os: linux-aarch64-a3-16
-            tests: tests/e2e/nightly/single_node/models/test_deepseek_v3_2_w8a8.py
+          # TODO: Replace deepseek3.2-exp with deepseek3.2 after nightly tests pass
+          # - name: deepseek3_2-exp-w8a8
+          #   os: linux-aarch64-a3-16
+          #   tests: tests/e2e/nightly/single_node/models/test_deepseek_v3_2_exp_w8a8.py
     uses: ./.github/workflows/_e2e_nightly_single_node.yaml
     with:
       vllm: v0.13.0

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/.github/workflows/pr_test_full.yaml

@@ -38,7 +38,7 @@ concurrency:
 
 jobs:
   changes:
-    runs-on: linux-aarch64-a2b3-0
+    runs-on: linux-aarch64-a2-0
     if: ${{ contains(github.event.pull_request.labels.*.name, 'ready') && contains(github.event.pull_request.labels.*.name, 'ready-for-test') }}
     outputs:
       e2e_tracker: ${{ steps.filter.outputs.e2e_tracker }}
@@ -80,6 +80,6 @@ jobs:
     uses: ./.github/workflows/_e2e_test.yaml
     with:
       vllm: ${{ matrix.vllm_version }}
-      runner: linux-aarch64-a2b3
+      runner: linux-aarch64-a2
       image: swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/cann:8.5.0-910b-ubuntu22.04-py3.11
       type: full

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/.github/workflows/pr_test_light.yaml

@@ -41,7 +41,7 @@ jobs:
     with:
       vllm: v0.13.0
   changes:
-    runs-on: linux-aarch64-a2b3-0
+    runs-on: linux-aarch64-a2-0
     outputs:
       e2e_tracker: ${{ steps.filter.outputs.e2e_tracker }}
       ut_tracker: ${{ steps.filter.outputs.ut_tracker }}
@@ -84,7 +84,7 @@ jobs:
     with:
       vllm: ${{ matrix.vllm_version }}
       runner: linux-amd64-cpu-16-hk
-      image: quay.nju.edu.cn/ascend/cann:8.5.0-910b-ubuntu22.04-py3.11
+      image: quay.nju.edu.cn/ascend/cann:8.2.rc2-910b-ubuntu22.04-py3.11
       type: pr
 
   e2e-light:
@@ -99,6 +99,6 @@ jobs:
     uses: ./.github/workflows/_e2e_test.yaml
     with:
       vllm: ${{ matrix.vllm_version }}
-      runner: linux-aarch64-a2b3
+      runner: linux-aarch64-a2
       image: swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/cann:8.5.0-910b-ubuntu22.04-py3.11
       type: light

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/.github/workflows/schedule_image_build_and_push.yaml

@@ -11,36 +11,12 @@
 #   - Publish when tag with v* (pep440 version)  ===>  vllm-ascend:v1.2.3 / vllm-ascend:v1.2.3rc1
 name: Image Build and Push
 on:
-  pull_request:
-    branches:
-      - 'releases/*'
-    paths:
-      - '.github/workflows/schedule_image_build_and_push.yaml'
-      - 'Dockerfile*'
-      - 'vllm_ascend/**'
-      - 'setup.py'
-      - 'pyproject.toml'
-      - 'requirements.txt'
-      - 'cmake/**'
-      - 'CMakeLists.txt'
-      - 'csrc/**'
-    types: [ labeled, synchronize ]
+  schedule:
+    # UTC+8: 8am, 12pm, 16pm, 22pm
+    - cron: '0 0,4,8,14 * * *'
   push:
-    # Publish image when tagging, the Dockerfile in tag will be build as tag image
-    branches:
-      - 'releases/*'
     tags:
       - 'v*'
-    paths:
-      - '.github/workflows/schedule_image_build_and_push.yaml'
-      - 'Dockerfile*'
-      - 'vllm_ascend/**'
-      - 'setup.py'
-      - 'pyproject.toml'
-      - 'requirements.txt'
-      - 'cmake/**'
-      - 'CMakeLists.txt'
-      - 'csrc/**'
   workflow_dispatch:
     inputs:
       tag:
@@ -49,13 +25,8 @@ on:
         default: main
         required: true
 
-concurrency:
-  group: ${{ github.workflow }}-${{ github.ref_name }}
-  cancel-in-progress: true
-
 jobs:
   image_build:
-    if: ${{ github.event_name == 'push' || contains(github.event.pull_request.labels.*.name, 'image-build') }}
     name: Image Build and Push
     strategy:
       matrix:
@@ -81,7 +52,7 @@ jobs:
       dockerfile: ${{ matrix.build_meta.dockerfile }}
       suffix: ${{ matrix.build_meta.suffix }}
       quay_username: ${{ vars.QUAY_USERNAME }}
-      should_push: ${{ github.repository_owner == 'vllm-project' && github.event_name == 'push'}}
+      should_push: ${{ github.repository_owner == 'vllm-project' }}
       workflow_dispatch_tag: ${{ inputs.tag }}
     secrets:
       QUAY_PASSWORD: ${{ secrets.QUAY_PASSWORD }}

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/.github/workflows/schedule_release_code_and_wheel.yml

@@ -122,13 +122,7 @@ jobs:
           --exclude "libascend*.so" \
           --exclude "libtorch*.so" \
           --exclude "libopapi.so" \
-          --exclude "liberror_manager.so" \
-          --exclude "libruntime.so" \
-          --exclude "libmmpa.so" \
-          --exclude "libops_base.so" \
-          --exclude "libopapi_math.so" \
-          --exclude "libunified_dlog.so" \
-          --exclude "liboptiling.so"
+          --exclude "liberror_manager.so"
         done
         rm -f dist/*.whl
         mv dist/repaired/*.whl dist/

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/.github/workflows/schedule_test_vllm_main.yaml

@@ -34,6 +34,6 @@ jobs:
     uses: ./.github/workflows/_e2e_test.yaml
     with:
       vllm: main
-      runner: linux-aarch64-a2b3
+      runner: linux-aarch64-a2
       image: swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/cann:8.5.0-910b-ubuntu22.04-py3.11
       type: full

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/Dockerfile

@@ -73,7 +73,7 @@ RUN apt-get update -y && \
     rm -rf /var/lib/apt/lists/*
 
 # Install modelscope (for fast download), ray (for multinode) and torch-npu post version
-RUN python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' && \
+RUN python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' 'torch-npu==2.8.0.post1' -i https://mirrors.huaweicloud.com/ascend/repos/pypi && \
     python3 -m pip cache purge
 
 RUN echo "export LD_PRELOAD=/usr/lib/$(uname -m)-linux-gnu/libjemalloc.so.2:$LD_PRELOAD" >> ~/.bashrc

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/Dockerfile.310p

@@ -58,7 +58,7 @@ RUN export PIP_EXTRA_INDEX_URL=https://mirrors.huaweicloud.com/ascend/repos/pypi
     python3 -m pip cache purge
 
 # Install modelscope (for fast download), ray (for multinode) and torch-npu post version
-RUN python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' && \
+RUN python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' 'torch-npu==2.8.0.post1' -i https://mirrors.huaweicloud.com/ascend/repos/pypi && \
     python3 -m pip cache purge
 
 RUN echo "export LD_PRELOAD=/usr/lib/$(uname -m)-linux-gnu/libjemalloc.so.2:$LD_PRELOAD" >> ~/.bashrc

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/Dockerfile.310p.openEuler

@@ -25,7 +25,7 @@ ENV SOC_VERSION=$SOC_VERSION \
     OMP_NUM_THREADS=1
 
 RUN yum update -y && \
-    yum install -y patch python3-pip git vim wget net-tools gcc gcc-c++ make cmake numactl-devel jemalloc && \
+    yum install -y python3-pip git vim wget net-tools gcc gcc-c++ make cmake numactl-devel jemalloc && \
     rm -rf /var/cache/yum
 
 RUN pip config set global.index-url ${PIP_INDEX_URL}
@@ -48,12 +48,13 @@ RUN export PIP_EXTRA_INDEX_URL=https://mirrors.huaweicloud.com/ascend/repos/pypi
     source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
     source /usr/local/Ascend/nnal/atb/set_env.sh && \
     export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/Ascend/ascend-toolkit/latest/`uname -i`-linux/devlib && \
+    export CPLUS_INCLUDE_PATH=$CPLUS_INCLUDE_PATH:/usr/include/c++/12:/usr/include/c++/12/`uname -i`-openEuler-linux && \
     export SOC_VERSION=ASCEND310P3 && \
     python3 -m pip install -v -e /vllm-workspace/vllm-ascend/ --extra-index https://download.pytorch.org/whl/cpu/ && \
     python3 -m pip cache purge
 
 # Install modelscope (for fast download), ray (for multinode) and torch-npu post version
-RUN python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' && \
+RUN python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' 'torch-npu==2.8.0.post1' -i https://mirrors.huaweicloud.com/ascend/repos/pypi && \
     python3 -m pip cache purge
 
 RUN echo "export LD_PRELOAD=/usr/lib64/libjemalloc.so.2:$LD_PRELOAD" >> ~/.bashrc

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/Dockerfile.a3

@@ -72,7 +72,7 @@ RUN apt-get update -y && \
     rm -rf /var/lib/apt/lists/*
 
 # Install modelscope (for fast download), ray (for multinode) and torch-npu post version
-RUN python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' && \
+RUN python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' 'torch-npu==2.8.0.post1' -i https://mirrors.huaweicloud.com/ascend/repos/pypi && \
     python3 -m pip cache purge
 
 RUN echo "export LD_PRELOAD=/usr/lib/$(uname -m)-linux-gnu/libjemalloc.so.2:$LD_PRELOAD" >> ~/.bashrc

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/Dockerfile.a3.openEuler

@@ -34,7 +34,7 @@ COPY . /vllm-workspace/vllm-ascend/
 SHELL ["/bin/bash", "-c"]
 
 RUN yum update -y && \
-    yum install -y patch git vim wget net-tools gcc gcc-c++ make cmake numactl-devel jemalloc && \
+    yum install -y git vim wget net-tools gcc gcc-c++ make cmake numactl-devel jemalloc && \
     git clone --depth 1 --branch ${MOONCAKE_TAG} https://github.com/kvcache-ai/Mooncake /vllm-workspace/Mooncake && \
     cp /vllm-workspace/vllm-ascend/tools/mooncake_installer.sh /vllm-workspace/Mooncake/ && \
     ARCH=$(uname -m) && \
@@ -62,6 +62,7 @@ RUN export PIP_EXTRA_INDEX_URL=https://mirrors.huaweicloud.com/ascend/repos/pypi
     source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
     source /usr/local/Ascend/nnal/atb/set_env.sh && \
     export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/Ascend/ascend-toolkit/latest/`uname -i`-linux/devlib && \
+    export CPLUS_INCLUDE_PATH=$CPLUS_INCLUDE_PATH:/usr/include/c++/12:/usr/include/c++/12/`uname -i`-openEuler-linux && \
     python3 -m pip install -v -e /vllm-workspace/vllm-ascend/ --extra-index https://download.pytorch.org/whl/cpu/ && \
     python3 -m pip cache purge
 
@@ -71,7 +72,7 @@ RUN yum update -y && \
     rm -rf /var/cache/yum/*
 
 # Install modelscope (for fast download), ray (for multinode) and torch-npu post version
-RUN python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' && \
+RUN python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' 'torch-npu==2.8.0.post1' -i https://mirrors.huaweicloud.com/ascend/repos/pypi && \
     python3 -m pip cache purge
 
 RUN echo "export LD_PRELOAD=/usr/lib64/libjemalloc.so.2:$LD_PRELOAD" >> ~/.bashrc

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/Dockerfile.openEuler

@@ -34,7 +34,7 @@ COPY . /vllm-workspace/vllm-ascend/
 SHELL ["/bin/bash", "-c"]
 
 RUN yum update -y && \
-    yum install -y patch git vim wget net-tools gcc gcc-c++ make cmake numactl-devel jemalloc && \
+    yum install -y git vim wget net-tools gcc gcc-c++ make cmake numactl-devel jemalloc && \
     git clone --depth 1 --branch ${MOONCAKE_TAG} https://github.com/kvcache-ai/Mooncake /vllm-workspace/Mooncake && \
     cp /vllm-workspace/vllm-ascend/tools/mooncake_installer.sh /vllm-workspace/Mooncake/ && \
     ARCH=$(uname -m) && \
@@ -62,6 +62,7 @@ RUN export PIP_EXTRA_INDEX_URL=https://mirrors.huaweicloud.com/ascend/repos/pypi
     source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
     source /usr/local/Ascend/nnal/atb/set_env.sh && \
     export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/Ascend/ascend-toolkit/latest/`uname -i`-linux/devlib && \
+    export CPLUS_INCLUDE_PATH=$CPLUS_INCLUDE_PATH:/usr/include/c++/12:/usr/include/c++/12/`uname -i`-openEuler-linux && \
     python3 -m pip install -v -e /vllm-workspace/vllm-ascend/ --extra-index https://download.pytorch.org/whl/cpu/ && \
     python3 -m pip cache purge
 
@@ -71,7 +72,7 @@ RUN yum update -y && \
     rm -rf /var/cache/yum/*
 
 # Install modelscope (for fast download), ray (for multinode) and torch-npu post version
-RUN python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' && \
+RUN python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' 'torch-npu==2.8.0.post1' -i https://mirrors.huaweicloud.com/ascend/repos/pypi && \
     python3 -m pip cache purge
 
 RUN echo "export LD_PRELOAD=/usr/lib64/libjemalloc.so.2:$LD_PRELOAD" >> ~/.bashrc

{vllm_ascend-0.13.0 → vllm_ascend-0.13.0rc2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: vllm_ascend
-Version: 0.13.0
+Version: 0.13.0rc2
 Summary: vLLM Ascend backend plugin
 Home-page: https://github.com/vllm-project/vllm-ascend
 Author: vLLM-Ascend team
@@ -38,9 +38,9 @@ Requires-Dist: compressed_tensors>=0.11.0
 Requires-Dist: msgpack
 Requires-Dist: quart
 Requires-Dist: numba
-Requires-Dist: torch-npu==2.8.0.post2
+Requires-Dist: torch-npu==2.8.0
 Requires-Dist: arctic-inference==0.1.1
-Requires-Dist: transformers<5,>=4.57.4
+Requires-Dist: transformers>=4.57.3
 Requires-Dist: fastapi<0.124.0
 Requires-Dist: triton-ascend==3.2.0
 Dynamic: author