Make CI work and use cache speed up e2e test

This patch make CI work and use cache speed up e2e test, including:
  1. Change push (post merge ci) and pull_request (pr ci) trigger branch to main
  2. Make mypy work by ignore base_communicator and clear unused deps
  3. Several improvements for vllm_ascend_test:
  - use cache (pip, ms, hf) speed up e2e test (25mins --> 5mins)
  - switch `git clone` command to `action/checkout` to speedup checkout and
  - Enable sv for pytest for better info dump
  - Remove network host to resole `docker: conflicting ontions: cannot attach both user-defined and non-user-definednetwork-modes`, which is a problem on docker 1.45 but not on 1.39.
  4. Adapt MLA decode optimizations: vllm-project/vllm@cabaf4e

Signed-off-by: Yikun Jiang <yikunkero@gmail.com>

.github/workflows/actionlint.yml

@@ -19,14 +19,14 @@ name: Lint GitHub Actions workflows
 on:
   push:
     branches:
-      - "*"
+      - "main"
     paths:
       - '.github/workflows/*.ya?ml'
       - '.github/workflows/actionlint.*'
       - '.github/workflows/matchers/actionlint.json'
   pull_request:
     branches:
-      - "*"
+      - "main"
     paths:
       - '.github/workflows/*.ya?ml'
       - '.github/workflows/actionlint.*'

.github/workflows/mypy.yaml

@@ -22,14 +22,14 @@ on:
   # but only for the main branch
   push:
     branches:
-      - "*"
+      - "main"
     paths:
       - '**/*.py'
       - '.github/workflows/mypy.yaml'
       - 'tools/mypy.sh'
   pull_request:
     branches:
-      - "*"
+      - "main"
     # This workflow is only relevant when one of the following files changes.
     # However, we have github configured to expect and require this workflow
     # to run and pass before github with auto-merge a pull request. Until github
@@ -54,23 +54,20 @@ jobs:
         python-version: ${{ matrix.python-version }}
     - name: Install dependencies
       run: |
-        python -m pip install --upgrade pip
-        pip install mypy==1.11.1
-        pip install types-setuptools
-        pip install types-PyYAML
-        pip install types-requests
-        pip install types-setuptools
-        sudo apt-get update
-        sudo apt-get install -y git gcc-12 g++-12 libnuma-dev
-        sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-12 10 --slave /usr/bin/g++ g++ /usr/bin/g++-12
-        pip install cmake>=3.26 wheel packaging ninja "setuptools-scm>=8" numpy
-    - name: Download and Install vllm cpu
-      # TODO (cmq): clone vllm-project/vllm:main instead of cosdt/vllm:apply_plugin
+        pip install -r requirements-dev.txt 
+
+    - name: Checkout vllm-project/vllm repo
+      uses: actions/checkout@v4
+      with:
+        repository: vllm-project/vllm
+        path: vllm-empty
+
+    - name: Install vllm-project/vllm from source
+      working-directory: vllm-empty
       run: |
-        git clone https://github.com/cosdt/vllm -b apply_plugin
-        cd vllm
-        pip install -r requirements-cpu.txt --extra-index-url https://download.pytorch.org/whl/cpu
-        VLLM_TARGET_DEVICE=cpu python setup.py install
+        pip install -r requirements-build.txt --extra-index-url https://download.pytorch.org/whl/cpu
+        VLLM_TARGET_DEVICE=empty pip install .
+
     - name: Mypy
       run: |
         echo "::add-matcher::.github/workflows/matchers/mypy.json"

.github/workflows/ruff.yml

@@ -22,15 +22,15 @@ on:
   # but only for the main branch
   push:
     branches:
-      - "*"
+      - "main"
     paths:
       - "**/*.py"
       - requirements-lint.txt
       - .github/workflows/matchers/ruff.json
       - .github/workflows/ruff.yml
   pull_request:
     branches:
-      - "*"
+      - "main"
 
 jobs:
   ruff:

.github/workflows/shellcheck.yml

@@ -19,13 +19,13 @@ name: Lint shell scripts
 on:
   push:
     branches:
-      - "*"
+      - "main"
     paths:
       - '**/*.sh'
       - '.github/workflows/shellcheck.yml'
   pull_request:
     branches:
-      - "*"
+      - "main"
     paths:
       - '**/*.sh'
       - '.github/workflows/shellcheck.yml'

.github/workflows/vllm_ascend_test.yaml

@@ -15,7 +15,7 @@
 # limitations under the License.
 #
 
-name: 'vllm_ascend_ut'
+name: 'e2e test'
 
 on:
   push:
@@ -42,19 +42,19 @@ defaults:
 
 jobs:
   test:
-    name: run ut for vllm_ascend
-    runs-on: npu-arm64  # actionlint-ignore: runner-label
+    name: vLLM Ascend test (self-host)
+    runs-on: ascend-arm64  # actionlint-ignore: runner-label
 
     container:
       image: quay.io/ascend/cann:8.0.rc3.beta1-910b-ubuntu22.04-py3.10
       volumes:
         - /usr/local/dcmi:/usr/local/dcmi
         - /usr/local/bin/npu-smi:/usr/local/bin/npu-smi
         - /usr/local/Ascend/driver/lib64/:/usr/local/Ascend/driver/lib64/
-        - /data/disk1/actions-runner/_work/_repo/vllm:/root/vllm
+        # Use self-host cache speed up pip and model download
+        - /home/action/actions-runner/_work/cache:/github/home/.cache/
       options: >-
-        --network host
-        --device /dev/davinci1
+        --device /dev/davinci6
         --device /dev/davinci_manager
         --device /dev/devmm_svm
         --device /dev/hisi_hdc
@@ -70,7 +70,7 @@ jobs:
           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
 
-      - name: Checkout
+      - name: Checkout vllm-project/vllm-ascend repo
         uses: actions/checkout@v4
 
       - name: Install system dependencies
@@ -82,28 +82,25 @@ jobs:
         run: |
           pip install -r requirements-dev.txt
 
-      # This is the base vllm master code cached in the container to speed up
-      # the test. There are two branch by default:
-      # - up-main: the latest code of vllm
-      # - apply_plugin: the code of vllm with the plugin applied
-      - name: copy vllm
-        run: |
-          cp -r /root/vllm /root/vllm-empty
+      - name: Checkout vllm-project/vllm repo
+        uses: actions/checkout@v4
+        with:
+          repository: vllm-project/vllm
+          path: ./vllm-empty
 
-      - name: Install vLLM from source
-        working-directory: /root/vllm-empty
+      - name: Install vllm-project/vllm from source
+        working-directory: ./vllm-empty
         run: |
-          git checkout up-main
           VLLM_TARGET_DEVICE=empty pip install -e .
 
-      - name: Install vllm_ascend
+      - name: Install vllm-project/vllm-ascend
         run: |
           pip install -e .
 
-      - name: Run test
+      - name: Run vllm-project/vllm-ascend test
         run: |
-          pytest tests
+          pytest -sv tests
 
-      - name: Run vLLM native test
+      - name: Run vllm-project/vllm test
         run: |
           bash tools/npu-vllm-test.sh

.github/workflows/yapf.yml

@@ -22,13 +22,13 @@ on:
   # but only for the main branch
   push:
     branches:
-      - "*"
+      - "main"
     paths:
       - "**/*.py"
       - .github/workflows/yapf.yml
   pull_request:
     branches:
-      - "*"
+      - "main"
     paths:
       - "**/*.py"
       - .github/workflows/yapf.yml

mypy.ini

@@ -8,3 +8,7 @@ ignore_missing_imports = True
 
 [mypy-transformers.*]
 ignore_missing_imports = True
+
+; Remove this after https://github.com/vllm-project/vllm/pull/11324 merged
+[mypy-vllm.distributed.device_communicators.base_communicator]
+ignore_missing_imports = True

tools/npu-vllm-test.sh

@@ -20,7 +20,7 @@
 
 set -o pipefail
 
-TEST_DIR="/root/vllm-empty/tests"
+TEST_DIR="./vllm-empty/tests"
 TEST_FILES=(
     test_sequence.py
     # test_utils.py

vllm_ascend/platform.py

@@ -100,7 +100,7 @@ def check_and_update_config(cls, vllm_config: VllmConfig) -> None:
 
     @classmethod
     def get_attn_backend_cls(cls, selected_backend, head_size, dtype,
-                             kv_cache_dtype, block_size, use_v1):
+                             kv_cache_dtype, block_size, use_v1, use_mla):
         return "vllm_ascend.attention.AscendAttentionBackend"
 
     @classmethod