init v0.23.0

Signed-off-by: Sun Ruoxi <sunruoxi@4paradigm.com>

tests/e2e/pull_request/four_card/test_qwen3_next.py

@@ -0,0 +1,80 @@
+#
+# 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.
+# Adapted from vllm/tests/basic_correctness/test_basic_correctness.py
+#
+import os
+from unittest.mock import patch
+
+from tests.e2e.conftest import VllmRunner
+
+
+def test_qwen3_next_distributed_mp_full_decode_only_tp4():
+    example_prompts = [
+        "Hello, my name is",
+    ] * 4
+    max_tokens = 5
+    with VllmRunner(
+        "Qwen/Qwen3-Next-80B-A3B-Instruct",
+        tensor_parallel_size=4,
+        max_model_len=4096,
+        gpu_memory_utilization=0.8,
+        distributed_executor_backend="mp",
+        compilation_config={"cudagraph_mode": "FULL_DECODE_ONLY", "cudagraph_capture_sizes": [1, 2, 4]},
+    ) as vllm_model:
+        vllm_model.generate_greedy(example_prompts, max_tokens)
+        del vllm_model
+
+
+@patch.dict(os.environ, {"VLLM_ASCEND_ENABLE_FLASHCOMM1": "1"})
+@patch.dict(os.environ, {"HCCL_BUFFSIZE": "1024"})
+def test_qwen3_next_w8a8dynamic_distributed_mp_flash_comm_tp4():
+    example_prompts = [
+        "Hello, my name is",
+    ] * 4
+    max_tokens = 5
+    with VllmRunner(
+        "vllm-ascend/Qwen3-Next-80B-A3B-Instruct-W8A8",
+        tensor_parallel_size=4,
+        max_model_len=4096,
+        gpu_memory_utilization=0.7,
+        distributed_executor_backend="mp",
+        enable_expert_parallel=True,
+        enforce_eager=True,
+        quantization="ascend",
+    ) as vllm_model:
+        vllm_model.generate_greedy(example_prompts, max_tokens)
+        del vllm_model
+
+
+@patch.dict(os.environ, {"HCCL_BUFFSIZE": "1024"})
+def test_qwen3_next_distributed_mp_graph_mode_tp4():
+    example_prompts = [
+        "Hello, my name is",
+    ] * 4
+    max_tokens = 5
+    with VllmRunner(
+        "Qwen/Qwen3-Next-80B-A3B-Instruct",
+        tensor_parallel_size=4,
+        max_model_len=4096,
+        gpu_memory_utilization=0.8,
+        distributed_executor_backend="mp",
+        enable_expert_parallel=True,
+        cudagraph_capture_sizes=[1, 2, 4],
+        enforce_eager=False,
+    ) as vllm_model:
+        vllm_model.generate_greedy(example_prompts, max_tokens)
+        del vllm_model