Enable nightly test and add qwen3 32b test case (#3370)

### What this PR does / why we need it?
This PR adds a nightly test case for qwen3_32b bf16
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
by running the case

- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0

---------

Signed-off-by: jiangyunfan1 <jiangyunfan1@h-partners.com>
Signed-off-by: wangli <wangli858794774@gmail.com>
Signed-off-by: Yikun Jiang <yikunkero@gmail.com>
Co-authored-by: wangli <wangli858794774@gmail.com>
Co-authored-by: Yikun Jiang <yikunkero@gmail.com>

.github/workflows/_e2e_nightly.yaml

@@ -0,0 +1,105 @@
+#
+# 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 nightly test'
+
+on:
+  workflow_call:
+    inputs:
+      vllm:
+        required: true
+        type: string
+      runner:
+        required: true
+        type: string
+      image:
+        required: false
+        type: string
+        default: "swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/cann:8.2.rc1-910b-ubuntu22.04-py3.11"
+      tests:
+        required: true
+        type: string
+
+# 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-nightly:
+    name: e2e-nightly
+    runs-on: ${{ inputs.runner }}
+    container:
+      image: ${{ inputs.image }}
+      env:
+        VLLM_USE_MODELSCOPE: True
+    steps:
+      - name: Check npu and CANN info
+        run: |
+          npu-smi info
+          cat /usr/local/Ascend/ascend-toolkit/latest/"$(uname -i)"-linux/ascend_toolkit_install.info
+
+      - name: Config mirrors
+        run: |
+          sed -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
+          apt-get update -y
+          apt install git -y
+          git config --global url."https://gh-proxy.test.osinfra.cn/https://github.com/".insteadOf https://github.com/
+
+      - 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 gcc g++ cmake libnuma-dev
+
+      - name: Checkout vllm-project/vllm repo
+        uses: actions/checkout@v4
+        with:
+          repository: vllm-project/vllm
+          ref: ${{ inputs.vllm }}
+          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
+        env:
+          PIP_EXTRA_INDEX_URL: https://mirrors.huaweicloud.com/ascend/repos/pypi
+        run: |
+          pip install -r requirements-dev.txt
+          pip install -v -e .
+
+      - name: Run vllm-project/vllm-ascend test
+        env:
+          VLLM_WORKER_MULTIPROC_METHOD: spawn
+          VLLM_USE_MODELSCOPE: True
+        run: |
+          # TODO: enable more tests
+          pytest -sv ${{ inputs.tests }}

.github/workflows/vllm_ascend_test_nightly.yaml

@@ -0,0 +1,58 @@
+#
+# 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: 'ascend test / nightly'
+
+on:
+  schedule:
+      # Run test at 24:00 Beijing time (UTC+8)
+      - cron: "0 16 * * *"
+  workflow_dispatch:
+  pull_request: 
+    branches:
+      - 'main'
+      - '*-dev'
+    paths:
+      - 'tests/e2e/nightly/**'
+      - '.github/workflows/vllm_ascend_test_nightly.yaml'
+
+# 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: ascend-nightly-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  qwen3-32b:
+    strategy:
+      matrix:
+        # should add A3 chip runner when available
+        os: [linux-aarch64-a2-4]
+    # Note (yikun): If CI resource are limited we can split job into two chain jobs
+    # only trigger e2e test after lint passed and the change is e2e related with pull request.
+    uses: ./.github/workflows/_e2e_nightly.yaml
+    with:
+      vllm: v0.11.0
+      runner: ${{ matrix.os }}
+      tests: tests/e2e/nightly/models/test_qwen3_32b.py

tests/e2e/conftest.py

@@ -110,10 +110,10 @@ class RemoteOpenAIServer:
 
     def __init__(self,
                  model: str,
-                 server_host: str,
-                 server_port: int,
                  vllm_serve_args: list[str],
                  *,
+                 server_host: str = "0.0.0.0",
+                 server_port: int = 8080,
                  env_dict: Optional[dict[str, str]] = None,
                  seed: Optional[int] = 0,
                  auto_port: bool = True,

tests/e2e/multi_node/test_multi_dp.py

@@ -30,9 +30,9 @@ def test_multi_dp(config: MultiNodeConfig) -> None:
 
     with RemoteOpenAIServer(
             model_name,
-            config.server_host,
-            config.server_port,
             server_args,
+            server_host=config.server_host,
+            server_port=config.server_port,
             env_dict=env_dict,
             auto_port=False,
             seed=1024,

tests/e2e/nightly/models/test_qwen3_32b.py

@@ -0,0 +1,71 @@
+# 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.
+#
+from typing import Any
+
+import openai
+import pytest
+
+from tests.e2e.conftest import RemoteOpenAIServer
+
+MODELS = [
+    "Qwen/Qwen3-32B",
+]
+
+TENSOR_PARALLELS = [4]
+
+prompts = [
+    "San Francisco is a",
+]
+
+api_keyword_args = {
+    "max_tokens": 10,
+}
+
+
+@pytest.mark.asyncio
+@pytest.mark.parametrize("model", MODELS)
+@pytest.mark.parametrize("tp_size", TENSOR_PARALLELS)
+async def test_models(model: str, tp_size: int) -> None:
+    env_dict = {
+        "TASK_QUEUE_ENABLE": "1",
+        "OMP_PROC_BIND": "false",
+        "HCCL_OP_EXPANSION_MODE": "AIV",
+        "PAGED_ATTENTION_MASK_LEN": "5500"
+    }
+    server_args = [
+        "--no-enable-prefix-caching", "--tensor-parallel-size",
+        str(tp_size), "--port", "20002", "--max-model-len", "36864",
+        "--max-num-batched-tokens", "36864", "--block-size", "128",
+        "--trust-remote-code", "--gpu-memory-utilization", "0.9",
+        "--additional-config", '{"enable_weight_nz_layout":true}'
+    ]
+    request_keyword_args: dict[str, Any] = {
+        **api_keyword_args,
+    }
+    with RemoteOpenAIServer(model,
+                            server_args,
+                            server_port=20002,
+                            env_dict=env_dict,
+                            auto_port=False) as server:
+        client = server.get_async_client()
+        batch = await client.completions.create(
+            model=model,
+            prompt=prompts,
+            **request_keyword_args,
+        )
+        choices: list[openai.types.CompletionChoice] = batch.choices
+        assert choices[0].text, "empty response"