[Refactor] Consolidate sequence normalization and enc-dec parsing (#33928)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2026-02-06 15:43:47 +00:00
committed by GitHub
parent 4707f7ebb4
commit cd8b405bd0
38 changed files with 1271 additions and 863 deletions
+3 -2
View File
@@ -28,10 +28,11 @@ from vllm.entrypoints.pooling.utils import (
encode_pooling_output_base64,
encode_pooling_output_float,
)
from vllm.inputs.data import EmbedsPrompt, TokensPrompt
from vllm.inputs.data import TokensPrompt
from vllm.logger import init_logger
from vllm.outputs import PoolingOutput, PoolingRequestOutput
from vllm.pooling_params import PoolingParams
from vllm.renderers.inputs import TokPrompt
from vllm.utils.async_utils import merge_async_iterators
from vllm.utils.collection_utils import chunk_list
from vllm.utils.serial_utils import EmbedDType, Endianness
@@ -369,7 +370,7 @@ class OpenAIServingEmbedding(OpenAIServing):
async def _create_single_prompt_generator(
self,
ctx: EmbeddingServeContext,
engine_prompt: TokensPrompt | EmbedsPrompt,
engine_prompt: TokPrompt,
pooling_params: PoolingParams,
trace_headers: Mapping[str, str] | None,
prompt_index: int,
+6 -6
View File
@@ -33,8 +33,11 @@ from vllm.entrypoints.pooling.utils import (
encode_pooling_output_base64,
encode_pooling_output_float,
)
from vllm.inputs import PromptType
from vllm.logger import init_logger
from vllm.outputs import PoolingRequestOutput
from vllm.renderers.inputs import TokPrompt
from vllm.renderers.inputs.preprocess import prompt_to_seq
from vllm.utils.async_utils import merge_async_iterators
from vllm.utils.serial_utils import EmbedDType, EncodingFormat, Endianness
@@ -91,6 +94,7 @@ class OpenAIServingPooling(OpenAIServing):
"dimensions is currently not supported"
)
engine_prompts: Sequence[PromptType | TokPrompt]
if is_io_processor_request:
if self.io_processor is None:
raise ValueError(
@@ -102,14 +106,10 @@ class OpenAIServingPooling(OpenAIServing):
validated_prompt = self.io_processor.parse_request(request)
engine_prompts = await self.io_processor.pre_process_async(
raw_prompts = await self.io_processor.pre_process_async(
prompt=validated_prompt, request_id=request_id
)
if not isinstance(engine_prompts, Sequence) or isinstance(
engine_prompts, (str, bytes, bytearray)
):
engine_prompts = [engine_prompts]
engine_prompts = prompt_to_seq(raw_prompts)
elif isinstance(request, PoolingChatRequest):
error_check_ret = self._validate_chat_template(
request_chat_template=request.chat_template,