Merge branch 'vllm-project:main' into main

Author: zhaozx-cn

.github/workflows/image_310p_openeuler.yml

@@ -25,6 +25,7 @@ on:
       - 'cmake/**'
       - 'CMakeLists.txt'
       - 'csrc/**'
+    types: [ labeled ]
   push:
     # Publish image when tagging, the Dockerfile in tag will be build as tag image
     branches:
@@ -43,6 +44,11 @@ on:
       - 'CMakeLists.txt'
       - 'csrc/**'
 
+# only cancel in-progress runs of the same workflow
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
 jobs:
   build:
     name: vllm-ascend image build
@@ -52,6 +58,7 @@ jobs:
           'ubuntu-latest' ||
           'ubuntu-24.04-arm'
       }}
+    if: ${{ contains(github.event.pull_request.labels.*.name, 'ready') && contains(github.event.pull_request.labels.*.name, 'ready-for-test') }}
     steps:
     - uses: actions/checkout@v4
       with:

.github/workflows/image_310p_ubuntu.yml

@@ -25,6 +25,7 @@ on:
       - 'cmake/**'
       - 'CMakeLists.txt'
       - 'csrc/**'
+    types: [ labeled ]
   push:
     # Publish image when tagging, the Dockerfile in tag will be build as tag image
     branches:
@@ -42,12 +43,18 @@ on:
       - 'cmake/**'
       - 'CMakeLists.txt'
       - 'csrc/**'
+
+# only cancel in-progress runs of the same workflow
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
 jobs:
 
   build:
     name: vllm-ascend image build
     runs-on: ubuntu-latest
-
+    if: ${{ contains(github.event.pull_request.labels.*.name, 'ready') && contains(github.event.pull_request.labels.*.name, 'ready-for-test') }}
     steps:
     - uses: actions/checkout@v4
       with:

.github/workflows/image_a3_openeuler.yml

@@ -25,6 +25,7 @@ on:
       - 'cmake/**'
       - 'CMakeLists.txt'
       - 'csrc/**'
+    types: [ labeled ]
   push:
     # Publish image when tagging, the Dockerfile in tag will be build as tag image
     branches:
@@ -43,6 +44,11 @@ on:
       - 'CMakeLists.txt'
       - 'csrc/**'
 
+# only cancel in-progress runs of the same workflow
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
 jobs:
   build:
     name: vllm-ascend image build
@@ -52,6 +58,7 @@ jobs:
           'ubuntu-latest' ||
           'ubuntu-24.04-arm'
       }}
+    if: ${{ contains(github.event.pull_request.labels.*.name, 'ready') && contains(github.event.pull_request.labels.*.name, 'ready-for-test') }}
     steps:
     - uses: actions/checkout@v4
       with:

.github/workflows/image_a3_ubuntu.yml

@@ -25,6 +25,7 @@ on:
       - 'cmake/**'
       - 'CMakeLists.txt'
       - 'csrc/**'
+    types: [ labeled ]
   push:
     # Publish image when tagging, the Dockerfile in tag will be build as tag image
     branches:
@@ -42,12 +43,18 @@ on:
       - 'cmake/**'
       - 'CMakeLists.txt'
       - 'csrc/**'
+
+# only cancel in-progress runs of the same workflow
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
 jobs:
 
   build:
     name: vllm-ascend image build
     runs-on: ubuntu-latest
-
+    if: ${{ contains(github.event.pull_request.labels.*.name, 'ready') && contains(github.event.pull_request.labels.*.name, 'ready-for-test') }}
     steps:
     - uses: actions/checkout@v4
       with:

.github/workflows/image_openeuler.yml

@@ -24,6 +24,7 @@ on:
       - 'cmake/**'
       - 'CMakeLists.txt'
       - 'csrc/**'
+    types: [ labeled ]
   push:
     # Publish image when tagging, the Dockerfile in tag will be build as tag image
     branches:
@@ -42,6 +43,11 @@ on:
       - 'CMakeLists.txt'
       - 'csrc/**'
 
+# only cancel in-progress runs of the same workflow
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
 jobs:
   build:
     name: vllm-ascend image build
@@ -51,6 +57,7 @@ jobs:
           'ubuntu-latest' ||
           'ubuntu-24.04-arm'
       }}
+    if: ${{ contains(github.event.pull_request.labels.*.name, 'ready') && contains(github.event.pull_request.labels.*.name, 'ready-for-test') }}
     steps:
     - uses: actions/checkout@v4
       with:

.github/workflows/image_ubuntu.yml

@@ -25,6 +25,7 @@ on:
       - 'cmake/**'
       - 'CMakeLists.txt'
       - 'csrc/**'
+    types: [ labeled ]
   push:
     # Publish image when tagging, the Dockerfile in tag will be build as tag image
     branches:
@@ -42,12 +43,18 @@ on:
       - 'cmake/**'
       - 'CMakeLists.txt'
       - 'csrc/**'
+
+# only cancel in-progress runs of the same workflow
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
 jobs:
 
   build:
     name: vllm-ascend image build
     runs-on: ubuntu-latest
-
+    if: ${{ contains(github.event.pull_request.labels.*.name, 'ready') && contains(github.event.pull_request.labels.*.name, 'ready-for-test') }}
     steps:
     - uses: actions/checkout@v4
       with:

.github/workflows/pre-commit.yml

@@ -8,19 +8,13 @@ permissions:
 
 jobs:
   pre-commit:
-    runs-on: linux-amd64-cpu-8
-    container:
-      image: swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/cann:8.2.rc1-910b-ubuntu22.04-py3.11
+    runs-on: ubuntu-latest
     steps:
-    - 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
-        apt-get update -y
-        apt install git -y
     - name: Checkout vllm-project/vllm-ascend repo
       uses: actions/checkout@v4
+    - uses: actions/setup-python@42375524e23c412d93fb67b49958b491fce71c38 # v5.4.0
+      with:
+        python-version: "3.11"
     - run: echo "::add-matcher::.github/workflows/matchers/actionlint.json"
     - run: echo "::add-matcher::.github/workflows/matchers/mypy.json"
     - name: Checkout vllm-project/vllm repo
@@ -36,9 +30,9 @@ jobs:
     - name: Install vllm-ascend dev
       run: |
         pip install -r requirements-dev.txt --extra-index-url https://download.pytorch.org/whl/cpu
-        git config --global --add safe.directory '*'
-    - name: Run pre-commit check
+    - uses: pre-commit/action@2c7b3805fd2a0fd8c1884dcaebf91fc102a13ecd # v3.0.1
       env:
-        SHELLCHECK_OPTS: "--exclude=SC2046,SC2006,SC2086"
-        GOPROXY: "https://goproxy.cn,direct"
-      run: pre-commit run --all-files --hook-stage manual
+        SHELLCHECK_OPTS: "--exclude=SC2046,SC2006,SC2086" # Exclude SC2046, SC2006, SC2086 for actionlint
+      with:
+        extra_args: --all-files --hook-stage manual
+

.github/workflows/vllm_ascend_test.yaml

@@ -43,9 +43,8 @@ jobs:
     uses: ./.github/workflows/pre-commit.yml
 
   changes:
-    runs-on: linux-amd64-cpu-0
-    container:
-      image: swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/cann:8.2.rc1-910b-ubuntu22.04-py3.11
+    if: github.event_name == 'pull_request'
+    runs-on: ubuntu-latest
     outputs:
       e2e_tracker: ${{ steps.filter.outputs.e2e_tracker }}
       ut_tracker: ${{ steps.filter.outputs.ut_tracker }}
@@ -258,4 +257,4 @@ jobs:
           VLLM_WORKER_MULTIPROC_METHOD: spawn
           VLLM_USE_MODELSCOPE: True
         run: |
-          pytest -sv tests/e2e/multicard/test_qwen3_moe.py::test_models_distributed_Qwen3_MOE_TP2_WITH_EP
+          pytest -sv tests/e2e/multicard/test_qwen3_moe.py::test_models_distributed_Qwen3_MOE_TP2_WITH_EP

.github/workflows/vllm_ascend_test_full.yaml

@@ -38,10 +38,8 @@ concurrency:
 
 jobs:
   changes:
-    runs-on: linux-amd64-cpu-0
+    runs-on: ubuntu-latest
     if: ${{ contains(github.event.pull_request.labels.*.name, 'ready') && contains(github.event.pull_request.labels.*.name, 'ready-for-test') }}
-    container:
-      image: swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/cann:8.2.rc1-910b-ubuntu22.04-py3.11
     outputs:
       e2e_tracker: ${{ steps.filter.outputs.e2e_tracker }}
       ut_tracker: ${{ steps.filter.outputs.ut_tracker }}
@@ -226,6 +224,8 @@ jobs:
           pytest -sv tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_Qwen3_W4A8DYNAMIC
           pytest -sv tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_DeepSeek_W4A8DYNAMIC
           pytest -sv tests/e2e/multicard/test_offline_inference_distributed.py::test_sp_for_qwen3_moe
+          pytest -sv tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_Qwen_Dense_with_flashcomm_v1
+          pytest -sv tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_Qwen_Dense_with_prefetch_mlp_weight
 
           #pytest -sv tests/e2e/multicard/test_pipeline_parallel.py
           pytest -sv tests/e2e/multicard/test_prefix_caching.py

tests/e2e/multicard/test_offline_inference_distributed.py

@@ -31,7 +31,9 @@
 
 os.environ["PYTORCH_NPU_ALLOC_CONF"] = "max_split_size_mb:256"
 
-QWEN_DENSE_MODELS = ["Qwen/QwQ-32B", "Qwen/Qwen-32B"]
+QWEN_DENSE_MODELS = [
+    "vllm-ascend/Qwen3-8B-W8A8", "vllm-ascend/Qwen2.5-0.5B-Instruct-W8A8"
+]
 
 
 def test_models_distributed_QwQ():
@@ -170,6 +172,29 @@ def test_models_distributed_Qwen_Dense_with_flashcomm_v1(model, enforce_eager):
             max_model_len=8192,
             enforce_eager=enforce_eager,
             dtype="auto",
-            tensor_parallel_size=4,
+            tensor_parallel_size=2,
+            quantization="ascend",
+    ) as vllm_model:
+        vllm_model.generate_greedy(example_prompts, max_tokens)
+
+
+@pytest.mark.parametrize("enforce_eager", [True, False])
+@pytest.mark.parametrize("model", QWEN_DENSE_MODELS)
+@patch.dict(os.environ, {"VLLM_ASCEND_ENABLE_DENSE_OPTIMIZE": "1"})
+@patch.dict(os.environ, {"VLLM_ASCEND_ENABLE_PREFETCH_MLP": "1"})
+def test_models_distributed_Qwen_Dense_with_prefetch_mlp_weight(
+        model, enforce_eager):
+    example_prompts = [
+        "Hello, my name is",
+    ]
+    max_tokens = 5
+
+    with VllmRunner(
+            snapshot_download(model),
+            max_model_len=8192,
+            enforce_eager=enforce_eager,
+            dtype="auto",
+            tensor_parallel_size=2,
+            quantization="ascend",
     ) as vllm_model:
         vllm_model.generate_greedy(example_prompts, max_tokens)