[Misc] Refactor additional_config (#1029)

More and more config options are added to additional_config. This PR
provide a new AscendConfig to manage these config options by an easier
way to make code cleaner and readable.

 This PR also added the `additional_config` doc for users.

Added the test_ascend_config.py to make sure the new AscendConfig works
as expect.

TODO: Add e2e test with torchair and deepseek once the CI resource is
available.

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>

tests/singlecard/test_aclgraph.py

@@ -0,0 +1,95 @@
+#
+# 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.
+#
+"""
+Compare the outputs of vLLM with and without aclgraph.
+
+Run `pytest tests/compile/test_aclgraph.py`.
+"""
+
+import os
+
+import pytest
+import torch
+from vllm import LLM, SamplingParams
+
+from tests.conftest import VllmRunner
+from tests.model_utils import check_outputs_equal
+
+MODELS = ["Qwen/Qwen2.5-0.5B-Instruct"]
+
+
+@pytest.mark.skipif(os.getenv("VLLM_USE_V1") == "0",
+                    reason="aclgraph only support on v1")
+@pytest.mark.parametrize("model", MODELS)
+@pytest.mark.parametrize("max_tokens", [32])
+def test_models(
+    model: str,
+    max_tokens: int,
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    with monkeypatch.context() as m:
+        prompts = [
+            "Hello, my name is", "The president of the United States is",
+            "The capital of France is", "The future of AI is"
+        ]
+
+        # aclgraph only support on v1
+        m.setenv("VLLM_USE_V1", "1")
+
+        sampling_params = SamplingParams(max_tokens=max_tokens,
+                                         temperature=0.0)
+        # TODO: change to use vllmrunner when the registry of custom op is solved
+        # while running pytest
+        vllm_model = LLM(model)
+        vllm_aclgraph_outputs = vllm_model.generate(prompts, sampling_params)
+        del vllm_model
+        torch.npu.empty_cache()
+
+        vllm_model = LLM(model, enforce_eager=True)
+        vllm_eager_outputs = vllm_model.generate(prompts, sampling_params)
+        del vllm_model
+        torch.npu.empty_cache()
+
+    vllm_aclgraph_outputs_list = []
+    for output in vllm_aclgraph_outputs:
+        vllm_aclgraph_outputs_list.append(
+            (output.outputs[0].index, output.outputs[0].text))
+
+    vllm_eager_outputs_list = []
+    for output in vllm_eager_outputs:
+        vllm_eager_outputs_list.append(
+            (output.outputs[0].index, output.outputs[0].text))
+
+    check_outputs_equal(
+        outputs_0_lst=vllm_eager_outputs_list,
+        outputs_1_lst=vllm_aclgraph_outputs_list,
+        name_0="vllm_eager_outputs",
+        name_1="vllm_aclgraph_outputs",
+    )
+
+
+@pytest.mark.skipif(os.getenv("VLLM_USE_V1") == "0",
+                    reason="aclgraph only support on v1")
+def test_deepseek_raises_error(monkeypatch: pytest.MonkeyPatch) -> None:
+    with monkeypatch.context() as m:
+        m.setenv("VLLM_USE_MODELSCOPE", "True")
+        m.setenv("VLLM_USE_V1", "1")
+        with pytest.raises(NotImplementedError) as excinfo:
+            VllmRunner("deepseek-ai/DeepSeek-V2-Lite-Chat",
+                       max_model_len=1024,
+                       enforce_eager=False)
+        assert "ACL Graph does not support deepseek" in str(excinfo.value)

tests/singlecard/test_ascend_config.py

@@ -0,0 +1,118 @@
+#
+# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
+# This file is a part of the vllm-ascend project.
+#
+# 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.
+import pytest
+
+from tests.conftest import VllmRunner
+from vllm_ascend.ascend_config import clear_ascend_config, get_ascend_config
+
+
+def _clean_up_ascend_config(func):
+
+    def wrapper(*args, **kwargs):
+        clear_ascend_config()
+        func(*args, **kwargs)
+        clear_ascend_config()
+
+    return wrapper
+
+
+@_clean_up_ascend_config
+def test_run_without_ascend_config():
+    with VllmRunner("facebook/opt-125m"):
+        ascend_config = get_ascend_config()
+
+        assert not ascend_config.torchair_graph_config.enabled
+        assert not ascend_config.torchair_graph_config.use_cached_graph
+        assert ascend_config.torchair_graph_config.graph_batch_sizes == []
+        assert not ascend_config.torchair_graph_config.graph_batch_sizes_init
+        assert not ascend_config.ascend_scheduler_config.enabled
+        assert ascend_config.expert_tensor_parallel_size == 1
+
+
+@_clean_up_ascend_config
+def test_run_with_ascend_config():
+    input_additional_config = {
+        "torchair_graph_config": {
+            # torchair graph only works with deepseek. The e2e test should be added
+            # in multicard test with deepseek models.
+            "enabled": False,
+            "use_cached_graph": True,
+            "graph_batch_sizes": [1, 2, 4, 8],
+            "graph_batch_sizes_init": False,
+        },
+        "ascend_scheduler_config": {
+            "enabled": True,
+            "enable_chunked_prefill": True,
+        },
+        "expert_tensor_parallel_size": 1
+    }
+    with VllmRunner("facebook/opt-125m",
+                    additional_config=input_additional_config):
+        ascend_config = get_ascend_config()
+
+        assert not ascend_config.torchair_graph_config.enabled
+        assert ascend_config.torchair_graph_config.use_cached_graph
+        assert ascend_config.torchair_graph_config.graph_batch_sizes == [
+            1, 2, 4, 8
+        ]
+        assert not ascend_config.torchair_graph_config.graph_batch_sizes_init
+        assert ascend_config.ascend_scheduler_config.enabled
+        assert ascend_config.ascend_scheduler_config.enable_chunked_prefill
+        assert ascend_config.expert_tensor_parallel_size == 1
+
+
+@_clean_up_ascend_config
+def test_ascend_config_init_error():
+    # ascend_config should be initialized first
+    with pytest.raises(RuntimeError):
+        _ = get_ascend_config()
+
+
+@_clean_up_ascend_config
+def test_ascend_config_load_error():
+    # graph_batch_sizes should be list.
+    with pytest.raises(TypeError):
+        input_additional_config_fake_1 = {
+            "torchair_graph_config": {
+                "graph_batch_sizes": "fake_size",
+            },
+        }
+        with VllmRunner("facebook/opt-125m",
+                        additional_config=input_additional_config_fake_1):
+            pass
+
+    # graph_batch_sizes_init should not be True when graph_batch_sizes is not empty.
+    with pytest.raises(ValueError):
+        input_additional_config_fake_2 = {
+            "torchair_graph_config": {
+                "graph_batch_sizes": [1, 2, 4, 8],
+                "graph_batch_sizes_init": True,
+            },
+        }
+        with VllmRunner("facebook/opt-125m",
+                        additional_config=input_additional_config_fake_2):
+            pass
+
+    # torchair graph only works with deepseek.
+    with pytest.raises(NotImplementedError):
+        input_additional_config_fake_2 = {
+            "torchair_graph_config": {
+                "enabled": True,
+            },
+        }
+        with VllmRunner("facebook/opt-125m",
+                        additional_config=input_additional_config_fake_2):
+            pass