biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Frontend] Delegate preprocessing to OpenAIServingRender (#36483)

Browse Source 
Signed-off-by: Sage Ahrac <sagiahrak@gmail.com>
This commit is contained in:
Sage
2026-03-13 09:39:43 +02:00
committed by GitHub
parent a4ad9db541
commit a2268617cf
10 changed files with 203 additions and 196 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
tests/entrypoints/openai/test_chat_error.py
View File

@@ -13,6 +13,7 @@ from vllm.entrypoints.openai.chat_completion.serving import OpenAIServingChat
from vllm.entrypoints.openai.engine.protocol import GenerationError
from vllm.entrypoints.openai.models.protocol import BaseModelPath
from vllm.entrypoints.openai.models.serving import OpenAIServingModels
from vllm.entrypoints.serve.render.serving import OpenAIServingRender
from vllm.outputs import CompletionOutput, RequestOutput
from vllm.renderers.hf import HfRenderer
from vllm.tokenizers.registry import tokenizer_args_from_config
@@ -84,10 +85,20 @@ def _build_serving_chat(engine: AsyncLLM) -> OpenAIServingChat:
engine_client=engine,
base_model_paths=BASE_MODEL_PATHS,
)
serving_render = OpenAIServingRender(
model_config=engine.model_config,
renderer=engine.renderer,
io_processor=engine.io_processor,
model_registry=models.registry,
request_logger=None,
chat_template=None,
chat_template_content_format="auto",
)
serving_chat = OpenAIServingChat(
engine,
models,
response_role="assistant",
openai_serving_render=serving_render,
request_logger=None,
chat_template=None,
chat_template_content_format="auto",
@@ -100,7 +111,9 @@ def _build_serving_chat(engine: AsyncLLM) -> OpenAIServingChat:
[{"prompt_token_ids": [1, 2, 3]}],
)
serving_chat._preprocess_chat = AsyncMock(side_effect=_fake_preprocess_chat)
serving_chat.openai_serving_render._preprocess_chat = AsyncMock(
side_effect=_fake_preprocess_chat
)
return serving_chat

11
tests/entrypoints/openai/test_completion_error.py
View File

@@ -13,6 +13,7 @@ from vllm.entrypoints.openai.completion.serving import OpenAIServingCompletion
from vllm.entrypoints.openai.engine.protocol import GenerationError
from vllm.entrypoints.openai.models.protocol import BaseModelPath
from vllm.entrypoints.openai.models.serving import OpenAIServingModels
from vllm.entrypoints.serve.render.serving import OpenAIServingRender
from vllm.outputs import CompletionOutput, RequestOutput
from vllm.renderers.hf import HfRenderer
from vllm.tokenizers.registry import tokenizer_args_from_config
@@ -74,9 +75,19 @@ def _build_serving_completion(engine: AsyncLLM) -> OpenAIServingCompletion:
engine_client=engine,
base_model_paths=BASE_MODEL_PATHS,
)
serving_render = OpenAIServingRender(
model_config=engine.model_config,
renderer=engine.renderer,
io_processor=engine.io_processor,
model_registry=models.registry,
request_logger=None,
chat_template=None,
chat_template_content_format="auto",
)
return OpenAIServingCompletion(
engine,
models,
openai_serving_render=serving_render,
request_logger=None,
)

12
tests/entrypoints/openai/test_lora_resolvers.py
View File

@@ -14,6 +14,7 @@ from vllm.entrypoints.openai.completion.serving import OpenAIServingCompletion
from vllm.entrypoints.openai.engine.protocol import ErrorResponse
from vllm.entrypoints.openai.models.protocol import BaseModelPath
from vllm.entrypoints.openai.models.serving import OpenAIServingModels
from vllm.entrypoints.serve.render.serving import OpenAIServingRender
from vllm.lora.request import LoRARequest
from vllm.lora.resolver import LoRAResolver, LoRAResolverRegistry
from vllm.renderers.hf import HfRenderer
@@ -145,8 +146,17 @@ def mock_serving_setup():
base_model_paths=BASE_MODEL_PATHS,
)
serving_render = OpenAIServingRender(
model_config=mock_engine.model_config,
renderer=mock_engine.renderer,
io_processor=mock_engine.io_processor,
model_registry=models.registry,
request_logger=None,
chat_template=None,
chat_template_content_format="auto",
)
serving_completion = OpenAIServingCompletion(
mock_engine, models, request_logger=None
mock_engine, models, openai_serving_render=serving_render, request_logger=None
)
return mock_engine, serving_completion

82
tests/entrypoints/openai/test_serving_chat.py
View File

@@ -21,8 +21,13 @@ from vllm.entrypoints.openai.engine.protocol import (
ErrorResponse,
RequestResponseMetadata,
)
from vllm.entrypoints.openai.models.serving import BaseModelPath, OpenAIServingModels
from vllm.entrypoints.openai.models.serving import (
BaseModelPath,
OpenAIModelRegistry,
OpenAIServingModels,
)
from vllm.entrypoints.openai.parser.harmony_utils import get_encoding
from vllm.entrypoints.serve.render.serving import OpenAIServingRender
from vllm.exceptions import VLLMValidationError
from vllm.inputs import TokensPrompt
from vllm.outputs import CompletionOutput, RequestOutput
@@ -557,15 +562,32 @@ def _build_renderer(model_config: MockModelConfig):
)
def _build_serving_render(
engine, model_registry: OpenAIModelRegistry
) -> OpenAIServingRender:
return OpenAIServingRender(
model_config=engine.model_config,
renderer=engine.renderer,
io_processor=engine.io_processor,
model_registry=model_registry,
request_logger=None,
chat_template=CHAT_TEMPLATE,
chat_template_content_format="auto",
)
def _build_serving_chat(engine: AsyncLLM) -> OpenAIServingChat:
models = OpenAIServingModels(
engine_client=engine,
base_model_paths=BASE_MODEL_PATHS,
)
openai_serving_render = _build_serving_render(engine, models.registry)
serving_chat = OpenAIServingChat(
engine,
models,
response_role="assistant",
openai_serving_render=openai_serving_render,
chat_template=CHAT_TEMPLATE,
chat_template_content_format="auto",
request_logger=None,
@@ -586,10 +608,13 @@ async def _async_serving_chat_init():
engine = MockEngine()
models = OpenAIServingModels(engine, BASE_MODEL_PATHS)
openai_serving_render = _build_serving_render(engine, models.registry)
serving_completion = OpenAIServingChat(
engine,
models,
response_role="assistant",
openai_serving_render=openai_serving_render,
chat_template=CHAT_TEMPLATE,
chat_template_content_format="auto",
request_logger=None,
@@ -1182,7 +1207,9 @@ class TestServingChatWithHarmony:
# Test the Harmony messages for the first turn's input
req = ChatCompletionRequest(model=MODEL_NAME, messages=messages)
input_messages, _ = serving_chat._make_request_with_harmony(req)
input_messages, _ = (
serving_chat.openai_serving_render._make_request_with_harmony(req)
)
verify_harmony_messages(
input_messages,
[
@@ -1209,7 +1236,9 @@ class TestServingChatWithHarmony:
# Test the Harmony messages for the second turn's input
req_2 = ChatCompletionRequest(model=MODEL_NAME, messages=messages)
input_messages_2, _ = serving_chat._make_request_with_harmony(req_2)
input_messages_2, _ = (
serving_chat.openai_serving_render._make_request_with_harmony(req_2)
)
verify_harmony_messages(
input_messages_2,
[
@@ -1230,7 +1259,9 @@ class TestServingChatWithHarmony:
# Test the Harmony messages for the first turn's input
req = ChatCompletionRequest(model=MODEL_NAME, messages=messages, tools=tools)
input_messages, _ = serving_chat._make_request_with_harmony(req)
input_messages, _ = (
serving_chat.openai_serving_render._make_request_with_harmony(req)
)
verify_harmony_messages(
input_messages,
[
@@ -1274,7 +1305,9 @@ class TestServingChatWithHarmony:
# Test the Harmony messages for the second turn's input
req_2 = ChatCompletionRequest(model=MODEL_NAME, messages=messages, tools=tools)
input_messages_2, _ = serving_chat._make_request_with_harmony(req_2)
input_messages_2, _ = (
serving_chat.openai_serving_render._make_request_with_harmony(req_2)
)
verify_harmony_messages(
input_messages_2,
[
@@ -1311,7 +1344,9 @@ class TestServingChatWithHarmony:
# Test the Harmony messages for the first turn's input
req = ChatCompletionRequest(model=MODEL_NAME, messages=messages, tools=tools)
input_messages, _ = serving_chat._make_request_with_harmony(req)
input_messages, _ = (
serving_chat.openai_serving_render._make_request_with_harmony(req)
)
verify_harmony_messages(
input_messages,
[
@@ -1355,7 +1390,9 @@ class TestServingChatWithHarmony:
# Test the Harmony messages for the second turn's input
req_2 = ChatCompletionRequest(model=MODEL_NAME, messages=messages, tools=tools)
input_messages_2, _ = serving_chat._make_request_with_harmony(req_2)
input_messages_2, _ = (
serving_chat.openai_serving_render._make_request_with_harmony(req_2)
)
verify_harmony_messages(
input_messages_2,
[
@@ -1392,7 +1429,9 @@ class TestServingChatWithHarmony:
# Test the Harmony messages for the first turn's input
req = ChatCompletionRequest(model=MODEL_NAME, messages=messages, tools=tools)
input_messages, _ = serving_chat._make_request_with_harmony(req)
input_messages, _ = (
serving_chat.openai_serving_render._make_request_with_harmony(req)
)
verify_harmony_messages(
input_messages,
[
@@ -1436,7 +1475,9 @@ class TestServingChatWithHarmony:
# Test the Harmony messages for the second turn's input
req_2 = ChatCompletionRequest(model=MODEL_NAME, messages=messages, tools=tools)
input_messages_2, _ = serving_chat._make_request_with_harmony(req_2)
input_messages_2, _ = (
serving_chat.openai_serving_render._make_request_with_harmony(req_2)
)
verify_harmony_messages(
input_messages_2,
[
@@ -1486,7 +1527,9 @@ class TestServingChatWithHarmony:
# Test the Harmony messages for the third turn's input
req_3 = ChatCompletionRequest(model=MODEL_NAME, messages=messages, tools=tools)
input_messages_3, _ = serving_chat._make_request_with_harmony(req_3)
input_messages_3, _ = (
serving_chat.openai_serving_render._make_request_with_harmony(req_3)
)
verify_harmony_messages(
input_messages_3,
[
@@ -1549,7 +1592,9 @@ class TestServingChatWithHarmony:
# Test the Harmony messages for the fourth turn's input
req_4 = ChatCompletionRequest(model=MODEL_NAME, messages=messages, tools=tools)
input_messages_4, _ = serving_chat._make_request_with_harmony(req_4)
input_messages_4, _ = (
serving_chat.openai_serving_render._make_request_with_harmony(req_4)
)
verify_harmony_messages(
input_messages_4,
[
@@ -1598,7 +1643,9 @@ class TestServingChatWithHarmony:
},
]
req = ChatCompletionRequest(model=MODEL_NAME, messages=messages)
input_messages, _ = serving_chat._make_request_with_harmony(req)
input_messages, _ = (
serving_chat.openai_serving_render._make_request_with_harmony(req)
)
verify_harmony_messages(
input_messages,
@@ -1629,7 +1676,9 @@ class TestServingChatWithHarmony:
},
]
req = ChatCompletionRequest(model=MODEL_NAME, messages=messages)
input_messages, _ = serving_chat._make_request_with_harmony(req)
input_messages, _ = (
serving_chat.openai_serving_render._make_request_with_harmony(req)
)
verify_harmony_messages(
input_messages,
@@ -1658,7 +1707,9 @@ class TestServingChatWithHarmony:
},
]
req = ChatCompletionRequest(model=MODEL_NAME, messages=messages)
input_messages, _ = serving_chat._make_request_with_harmony(req)
input_messages, _ = (
serving_chat.openai_serving_render._make_request_with_harmony(req)
)
verify_harmony_messages(
input_messages,
@@ -1689,11 +1740,14 @@ async def test_tool_choice_validation_without_parser():
engine_client=mock_engine,
base_model_paths=BASE_MODEL_PATHS,
)
openai_serving_render = _build_serving_render(mock_engine, models.registry)
# Create serving_chat without tool_parser (enable_auto_tools=False)
serving_chat = OpenAIServingChat(
mock_engine,
models,
response_role="assistant",
openai_serving_render=openai_serving_render,
chat_template=CHAT_TEMPLATE,
chat_template_content_format="auto",
request_logger=None,