[CI] Move applicable tests to CPU (#24080)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>

tests/tool_use/mistral/conftest.py

@@ -0,0 +1,41 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+import pytest
+import pytest_asyncio
+from huggingface_hub import snapshot_download
+
+from tests.utils import RemoteOpenAIServer
+from vllm.platforms import current_platform
+
+from .utils import ARGS, CONFIGS, ServerConfig
+
+
+# for each server config, download the model and return the config
+@pytest.fixture(scope="package", params=CONFIGS.keys())
+def server_config(request):
+    config = CONFIGS[request.param]
+
+    if current_platform.is_rocm() and not config.get("supports_rocm", True):
+        pytest.skip("The {} model can't be tested on the ROCm platform".format(
+            config["model"]))
+
+    # download model and tokenizer using transformers
+    snapshot_download(config["model"])
+    yield CONFIGS[request.param]
+
+
+# run this for each server config
+@pytest.fixture(scope="package")
+def server(request, server_config: ServerConfig):
+    model = server_config["model"]
+    args_for_model = server_config["arguments"]
+    with RemoteOpenAIServer(model, ARGS + args_for_model,
+                            max_wait_seconds=480) as server:
+        yield server
+
+
+@pytest_asyncio.fixture
+async def client(server: RemoteOpenAIServer):
+    async with server.get_async_client() as async_client:
+        yield async_client

tests/tool_use/mistral/test_mistral_tool_calls.py

@@ -0,0 +1,30 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+import openai
+import pytest
+
+from tests.tool_use.utils import MESSAGES_ASKING_FOR_TOOLS, WEATHER_TOOL
+
+
+# test: a tool_choice with mistral-tokenizer results in an ID of length 9
+@pytest.mark.asyncio
+async def test_tool_call_with_tool_choice(client: openai.AsyncOpenAI):
+    models = await client.models.list()
+    model_name: str = models.data[0].id
+    chat_completion = await client.chat.completions.create(
+        messages=MESSAGES_ASKING_FOR_TOOLS,
+        temperature=0,
+        max_completion_tokens=100,
+        model=model_name,
+        tools=[WEATHER_TOOL],
+        tool_choice=WEATHER_TOOL,
+        logprobs=False)
+
+    choice = chat_completion.choices[0]
+
+    assert choice.finish_reason != "tool_calls"  # "stop" or "length"
+    assert choice.message.role == "assistant"
+    assert choice.message.tool_calls is None \
+           or len(choice.message.tool_calls) == 1
+    assert len(choice.message.tool_calls[0].id) == 9  # length of 9 for mistral

tests/tool_use/mistral/utils.py

@@ -0,0 +1,34 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+from typing import Optional
+
+from typing_extensions import TypedDict
+
+
+class ServerConfig(TypedDict, total=False):
+    model: str
+    arguments: list[str]
+    system_prompt: Optional[str]
+    supports_parallel: Optional[bool]
+    supports_rocm: Optional[bool]
+
+
+ARGS: list[str] = ["--max-model-len", "1024"]
+
+CONFIGS: dict[str, ServerConfig] = {
+    "mistral": {
+        "model":
+        "mistralai/Mistral-7B-Instruct-v0.3",
+        "arguments": [
+            "--tokenizer-mode", "mistral",
+            "--ignore-patterns=\"consolidated.safetensors\""
+        ],
+        "system_prompt":
+        "You are a helpful assistant with access to tools. If a tool"
+        " that you have would be helpful to answer a user query, "
+        "call the tool. Otherwise, answer the user's query directly "
+        "without calling a tool. DO NOT CALL A TOOL THAT IS IRRELEVANT "
+        "to the user's question - just respond to it normally."
+    },
+}

tests/tool_use/test_glm4_moe_tool_parser.py

@@ -10,6 +10,8 @@ from vllm.entrypoints.openai.protocol import FunctionCall, ToolCall
 from vllm.entrypoints.openai.tool_parsers import Glm4MoeModelToolParser
 from vllm.transformers_utils.tokenizer import get_tokenizer
 
+pytestmark = pytest.mark.cpu_test
+
 pytest.skip("skip glm4_moe parser test", allow_module_level=True)
 # Use a common model that is likely to be available
 MODEL = "zai-org/GLM-4.5"

tests/tool_use/test_jamba_tool_parser.py

@@ -15,6 +15,8 @@ from vllm.entrypoints.openai.tool_parsers import JambaToolParser
 from vllm.transformers_utils.detokenizer_utils import detokenize_incrementally
 from vllm.transformers_utils.tokenizer import AnyTokenizer, get_tokenizer
 
+pytestmark = pytest.mark.cpu_test
+
 MODEL = "ai21labs/Jamba-tiny-dev"
 
 

tests/tool_use/test_kimi_k2_tool_parser.py

@@ -10,6 +10,8 @@ from vllm.entrypoints.openai.protocol import FunctionCall, ToolCall
 from vllm.entrypoints.openai.tool_parsers import KimiK2ToolParser
 from vllm.transformers_utils.tokenizer import get_tokenizer
 
+pytestmark = pytest.mark.cpu_test
+
 # Use a common model that is likely to be available
 MODEL = "moonshotai/Kimi-K2-Instruct"
 

tests/tool_use/test_minimax_tool_parser.py

@@ -12,6 +12,8 @@ from vllm.entrypoints.openai.protocol import (ChatCompletionToolsParam,
 from vllm.entrypoints.openai.tool_parsers import MinimaxToolParser
 from vllm.transformers_utils.tokenizer import get_tokenizer
 
+pytestmark = pytest.mark.cpu_test
+
 # Use a common model that is likely to be available
 MODEL = "MiniMaxAi/MiniMax-M1-40k"
 

tests/tool_use/test_qwen3coder_tool_parser.py

@@ -18,6 +18,8 @@ from vllm.entrypoints.openai.tool_parsers.qwen3xml_tool_parser import (
 from vllm.transformers_utils.detokenizer_utils import detokenize_incrementally
 from vllm.transformers_utils.tokenizer import AnyTokenizer, get_tokenizer
 
+pytestmark = pytest.mark.cpu_test
+
 MODEL = "Qwen/Qwen3-Coder-30B-A3B-Instruct-FP8"
 
 

tests/tool_use/test_seed_oss_tool_parser.py

@@ -16,6 +16,8 @@ from vllm.entrypoints.openai.tool_parsers import SeedOssToolParser
 from vllm.transformers_utils.detokenizer_utils import detokenize_incrementally
 from vllm.transformers_utils.tokenizer import AnyTokenizer, get_tokenizer
 
+pytestmark = pytest.mark.cpu_test
+
 # Use a common model that is likely to be available
 MODEL = "ByteDance-Seed/Seed-OSS-36B-Instruct"
 

tests/tool_use/test_tool_choice_required.py

@@ -12,6 +12,8 @@ from vllm.entrypoints.openai.protocol import (ChatCompletionRequest,
                                               ChatCompletionToolsParam)
 from vllm.entrypoints.openai.serving_chat import OpenAIServingChat
 
+pytestmark = pytest.mark.cpu_test
+
 EXAMPLE_TOOLS = [
     {
         "type": "function",

tests/tool_use/test_xlam_tool_parser.py

@@ -14,6 +14,8 @@ from vllm.entrypoints.openai.tool_parsers import xLAMToolParser
 from vllm.transformers_utils.detokenizer_utils import detokenize_incrementally
 from vllm.transformers_utils.tokenizer import AnyTokenizer, get_tokenizer
 
+pytestmark = pytest.mark.cpu_test
+
 # Use a common model that is likely to be available
 MODEL = "Salesforce/Llama-xLAM-2-8B-fc-r"