[Frontend][Doc][5/N] Improve all pooling task | Polish encode (pooling) api & Document. (#25524)

Signed-off-by: wang.yuqi <noooop@126.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com>
This commit is contained in:
wang.yuqi
2025-10-30 12:13:05 +00:00
committed by GitHub
co-authored by gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> Cyrus Leung
parent 74374386e2
commit 4464723f22
27 changed files with 499 additions and 131 deletions
+2 -6
View File
@@ -107,6 +107,7 @@ from vllm.entrypoints.utils import (
)
from vllm.logger import init_logger
from vllm.reasoning import ReasoningParserManager
from vllm.tasks import POOLING_TASKS
from vllm.usage.usage_lib import UsageContext
from vllm.utils.argparse_utils import FlexibleArgumentParser
from vllm.utils.network_utils import is_valid_ipv6_address
@@ -1748,12 +1749,7 @@ async def init_app_state(
log_error_stack=args.log_error_stack,
)
)
if (
any(
task in supported_tasks
for task in ["token_embed", "token_classify", "plugin"]
)
)
if any(task in POOLING_TASKS for task in supported_tasks)
else None
)
state.openai_serving_embedding = (
+100 -8
View File
@@ -49,6 +49,8 @@ from openai.types.responses.response_reasoning_item import (
)
from openai_harmony import Message as OpenAIHarmonyMessage
from vllm.config.pooler import get_use_activation
from vllm.tasks import PoolingTask
from vllm.utils.serial_utils import (
EmbedDType,
EncodingFormat,
@@ -1669,8 +1671,58 @@ class EmbeddingChatRequest(OpenAIBaseModel):
EmbeddingRequest: TypeAlias = EmbeddingCompletionRequest | EmbeddingChatRequest
PoolingCompletionRequest = EmbeddingCompletionRequest
PoolingChatRequest = EmbeddingChatRequest
class PoolingCompletionRequest(EmbeddingCompletionRequest):
task: PoolingTask | None = None
softmax: bool | None = Field(
default=None,
description="softmax will be deprecated, please use use_activation instead.",
)
activation: bool | None = Field(
default=None,
description="activation will be deprecated, please use use_activation instead.",
)
use_activation: bool | None = Field(
default=None,
description="Whether to use activation for classification outputs. "
"If it is a classify or token_classify task, the default is True; "
"for other tasks, this value should be None.",
)
def to_pooling_params(self):
return PoolingParams(
truncate_prompt_tokens=self.truncate_prompt_tokens,
dimensions=self.dimensions,
normalize=self.normalize,
use_activation=get_use_activation(self),
)
class PoolingChatRequest(EmbeddingChatRequest):
task: PoolingTask | None = None
softmax: bool | None = Field(
default=None,
description="softmax will be deprecated, please use use_activation instead.",
)
activation: bool | None = Field(
default=None,
description="activation will be deprecated, please use use_activation instead.",
)
use_activation: bool | None = Field(
default=None,
description="Whether to use activation for classification outputs. "
"If it is a classify or token_classify task, the default is True; "
"for other tasks, this value should be None.",
)
def to_pooling_params(self):
return PoolingParams(
truncate_prompt_tokens=self.truncate_prompt_tokens,
dimensions=self.dimensions,
normalize=self.normalize,
use_activation=get_use_activation(self),
)
T = TypeVar("T")
@@ -1686,6 +1738,7 @@ class IOProcessorRequest(OpenAIBaseModel, Generic[T]):
"""
data: T
task: PoolingTask = "plugin"
encoding_format: EncodingFormat = "float"
embed_dtype: EmbedDType = Field(
default="float32",
@@ -1749,14 +1802,27 @@ class ScoreRequest(OpenAIBaseModel):
),
)
activation: bool | None = None
softmax: bool | None = Field(
default=None,
description="softmax will be deprecated, please use use_activation instead.",
)
activation: bool | None = Field(
default=None,
description="activation will be deprecated, please use use_activation instead.",
)
use_activation: bool | None = Field(
default=None,
description="Whether to use activation for classification outputs. "
"Default is True.",
)
# --8<-- [end:score-extra-params]
def to_pooling_params(self):
return PoolingParams(
truncate_prompt_tokens=self.truncate_prompt_tokens,
activation=self.activation,
use_activation=get_use_activation(self),
)
@@ -1783,14 +1849,27 @@ class RerankRequest(OpenAIBaseModel):
),
)
activation: bool | None = None
softmax: bool | None = Field(
default=None,
description="softmax will be deprecated, please use use_activation instead.",
)
activation: bool | None = Field(
default=None,
description="activation will be deprecated, please use use_activation instead.",
)
use_activation: bool | None = Field(
default=None,
description="Whether to use activation for classification outputs. "
"Default is True.",
)
# --8<-- [end:rerank-extra-params]
def to_pooling_params(self):
return PoolingParams(
truncate_prompt_tokens=self.truncate_prompt_tokens,
activation=self.activation,
use_activation=get_use_activation(self),
)
@@ -1958,14 +2037,27 @@ class ClassificationRequest(OpenAIBaseModel):
),
)
activation: bool | None = None
softmax: bool | None = Field(
default=None,
description="softmax will be deprecated, please use use_activation instead.",
)
activation: bool | None = Field(
default=None,
description="activation will be deprecated, please use use_activation instead.",
)
use_activation: bool | None = Field(
default=None,
description="Whether to use activation for classification outputs. "
"Default is True.",
)
# --8<-- [end:classification-extra-params]
def to_pooling_params(self):
return PoolingParams(
truncate_prompt_tokens=self.truncate_prompt_tokens,
activation=self.activation,
use_activation=get_use_activation(self),
)
+16 -7
View File
@@ -170,15 +170,24 @@ class OpenAIServingPooling(OpenAIServing):
pooling_params = request.to_pooling_params()
pooling_task: PoolingTask
if "token_embed" in self.supported_tasks:
pooling_task = "token_embed"
elif "token_classify" in self.supported_tasks:
pooling_task = "token_classify"
elif "plugin" in self.supported_tasks:
pooling_task = "plugin"
if request.task is None:
if "token_embed" in self.supported_tasks:
pooling_task = "token_embed"
elif "token_classify" in self.supported_tasks:
pooling_task = "token_classify"
elif "plugin" in self.supported_tasks:
pooling_task = "plugin"
else:
return self.create_error_response(
f"pooling_task must be one of {self.supported_tasks}."
)
else:
pooling_task = request.task
if pooling_task not in self.supported_tasks:
return self.create_error_response(
f"pooling_task must be one of {self.supported_tasks}."
f"Task {pooling_task} is not supported, it"
f" must be one of {self.supported_tasks}."
)
try: