[CI]Add e2e test for 310p (#1879)

### What this PR does / why we need it?
Add e2e test for 310p:
trigger conditions：tag, labels(ready-for-test, e2e-310p-test), schedule
image: m.daocloud.io/quay.io/ascend/cann:8.1.rc1-310p-ubuntu22.04-py3.10
runner: linux-aarch64-310p-1, linux-aarch64-310p-4
model: IntervitensInc/pangu-pro-moe-model, Qwen/Qwen3-0.6B-Base,
Qwen/Qwen2.5-7B-Instruct

- vLLM version: v0.10.0
- vLLM main:
b917da442b

Signed-off-by: hfadzxy <starmoon_zhang@163.com>

.github/actionlint.yaml

@@ -7,4 +7,7 @@ self-hosted-runner:
     - linux-aarch64-a2-4
     - linux-aarch64-a2-8
     - linux-arm64-npu-static-8
+    - linux-aarch64-310p-1
+    - linux-aarch64-310p-2
+    - linux-aarch64-310p-4
     - ubuntu-24.04-arm

.github/workflows/vllm_ascend_test.yaml

@@ -203,7 +203,8 @@ jobs:
           --ignore=tests/e2e/singlecard/test_camem.py \
           --ignore=tests/e2e/singlecard/test_embedding.py \
           --ignore=tests/e2e/singlecard/spec_decode_v1/test_v1_mtp_correctness.py \
-          --ignore=tests/e2e/singlecard/spec_decode_v1/test_v1_spec_decode.py
+          --ignore=tests/e2e/singlecard/spec_decode_v1/test_v1_spec_decode.py \
+          --ignore=tests/e2e/singlecard/test_offline_inference_310p.py
           # ------------------------------------ v1 spec decode test ------------------------------------ #
           VLLM_USE_MODELSCOPE=True pytest -sv tests/e2e/singlecard/spec_decode_v1/test_v1_mtp_correctness.py
           # TODO: revert me when test_v1_spec_decode.py::test_ngram_correctness is fixed
@@ -280,4 +281,5 @@ jobs:
           pytest -sv tests/e2e/multicard/test_data_parallel.py
           pytest -sv tests/e2e/multicard/ --ignore=tests/e2e/multicard/test_ilama_lora_tp2.py \
             --ignore=tests/e2e/multicard/test_offline_inference_distributed.py \
-            --ignore=tests/e2e/multicard/test_data_parallel.py
+            --ignore=tests/e2e/multicard/test_data_parallel.py \
+            --ignore=tests/e2e/multicard/test_offline_inference_310p.py

.github/workflows/vllm_ascend_test_310p.yaml

@@ -0,0 +1,117 @@
+#
+# 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: 'e2e test / 310p-test'
+
+on:
+  push:
+    tags:
+      - 'v*'
+  schedule:
+    # Runs every 6 hours
+    - cron:  '0 */6 * * *'
+  pull_request:
+    types: [ labeled ]
+
+# 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}
+
+# only cancel in-progress runs of the same workflow
+# and ignore the lint / 1 card / 4 cards test type
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  e2e:
+    # e2e-310p-test will be triggered when tag 'e2e-310p-test' & 'ready-for-test' or schedule job
+    if: >- 
+      ${{ 
+        (contains(github.event.pull_request.labels.*.name, 'e2e-310p-test'))  && 
+        contains(github.event.pull_request.labels.*.name, 'ready-for-test') || 
+        github.event_name == 'schedule' || github.event_name == 'push' 
+        }}
+    strategy:
+      max-parallel: 2
+      matrix:
+        os: [linux-aarch64-310p-1, linux-aarch64-310p-4]
+        vllm_version: [main, v0.10.0]
+    name: 310p e2e test
+    runs-on: ${{ matrix.os }}
+    container:
+      # TODO(yikun): Remove m.daocloud.io prefix when infra proxy ready
+      image: swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/cann:8.1.rc1-310p-ubuntu22.04-py3.10
+      env:
+        VLLM_LOGGING_LEVEL: ERROR
+        VLLM_USE_MODELSCOPE: True
+    steps:        
+      - name: Check npu and CANN info
+        run: |
+          npu-smi info
+          cat /usr/local/Ascend/ascend-toolkit/latest/"$(uname -i)"-linux/ascend_toolkit_install.info
+
+      - name: Config mirrors
+        run: |
+          sed -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
+
+      - name: Install system dependencies
+        run: |
+          apt-get -y install `cat packages.txt`
+          apt-get -y install git vim wget net-tools gcc g++ cmake libnuma-dev curl gnupg2
+
+      - name: Checkout vllm-project/vllm repo
+        uses: actions/checkout@v4
+        with:
+          repository: vllm-project/vllm
+          ref: ${{ matrix.vllm_version }}
+          path: ./vllm-empty
+
+      - name: Install vllm-project/vllm from source
+        working-directory: ./vllm-empty
+        run: |
+          VLLM_TARGET_DEVICE=empty pip install -e .
+
+      - name: Install vllm-project/vllm-ascend
+        run: |
+          export PIP_EXTRA_INDEX_URL=https://mirrors.huaweicloud.com/ascend/repos/pypi
+          export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/Ascend/ascend-toolkit/latest/x86_64-linux/devlib
+          export SOC_VERSION=ASCEND310P3 
+          pip install -r requirements-dev.txt
+          pip install -v -e .
+
+      - name: Run e2e test
+        env:
+          VLLM_WORKER_MULTIPROC_METHOD: spawn
+          VLLM_USE_MODELSCOPE: True
+          PYTORCH_NPU_ALLOC_CONF: max_split_size_mb:256
+        run: |
+          if [[ "${{ matrix.os }}" == "linux-aarch64-310p-1" ]]; then
+            pytest -sv tests/e2e/singlecard/test_offline_inference_310p.py
+          else
+            pytest -sv tests/e2e/multicard/test_offline_inference_310p.py
+          fi

tests/e2e/multicard/test_offline_inference_310p.py

@@ -0,0 +1,62 @@
+#
+# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
+# Copyright 2023 The vLLM team.
+#
+# 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.
+import pytest
+import vllm  # noqa: F401
+
+import vllm_ascend  # noqa: F401
+from tests.e2e.conftest import VllmRunner
+
+# Pangu local model path
+MODELS = [
+    "IntervitensInc/pangu-pro-moe-model",
+]
+# set additional config for ascend scheduler and torchair graph
+ADDITIONAL_CONFIG = [{
+    "additional_config": {
+        "torchair_graph_config": {
+            "enabled": True
+        },
+        "ascend_scheduler_config": {
+            "enabled": True,
+        }
+    }
+}]
+
+
+@pytest.mark.parametrize("model", MODELS)
+@pytest.mark.parametrize("dtype", ["float16"])
+@pytest.mark.parametrize("max_tokens", [5])
+@pytest.mark.parametrize("enfore_eager", [True, False])
+@pytest.mark.parametrize("additional_config", ADDITIONAL_CONFIG)
+def test_pangu_model(model: str, dtype: str, max_tokens: int,
+                     enfore_eager: bool, additional_config: dict) -> None:
+    if enfore_eager:
+        additional_config = {}
+    example_prompts = [
+        "Hello, my name is",
+        "The future of AI is",
+    ]
+
+    with VllmRunner(model,
+                    tensor_parallel_size=4,
+                    dtype=dtype,
+                    max_model_len=1024,
+                    enforce_eager=True,
+                    enable_expert_parallel=True,
+                    additional_config=additional_config,
+                    distributed_executor_backend="mp") as vllm_model:
+        vllm_model.generate_greedy(example_prompts, max_tokens)

tests/e2e/singlecard/test_offline_inference_310p.py

@@ -0,0 +1,44 @@
+#
+# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
+# Copyright 2023 The vLLM team.
+#
+# 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.
+import pytest
+import vllm  # noqa: F401
+
+import vllm_ascend  # noqa: F401
+from tests.e2e.conftest import VllmRunner
+
+MODELS = ["Qwen/Qwen3-0.6B-Base", "Qwen/Qwen2.5-7B-Instruct"]
+
+
+@pytest.mark.parametrize("model", MODELS)
+@pytest.mark.parametrize("dtype", ["float16"])
+@pytest.mark.parametrize("max_tokens", [5])
+def test_models(model: str, dtype: str, max_tokens: int) -> None:
+    example_prompts = [
+        "Hello, my name is",
+        "The future of AI is",
+    ]
+
+    with VllmRunner(model,
+                    tensor_parallel_size=1,
+                    dtype=dtype,
+                    max_model_len=2048,
+                    enforce_eager=True,
+                    compilation_config={
+                        "custom_ops":
+                        ["none", "+rms_norm", "+rotary_embedding"]
+                    }) as vllm_model:
+        vllm_model.generate_greedy(example_prompts, max_tokens)