[Frontend] Refine the entrypoint class's inheritance hierarchy. (#47498)

Signed-off-by: wang.yuqi <yuqi.wang@daocloud.io>
This commit is contained in:
wang.yuqi
2026-07-03 10:50:06 +00:00
committed by GitHub
parent 18f658bb31
commit a14f57a3ac
23 changed files with 90 additions and 191 deletions
@@ -16,13 +16,13 @@ from transformers import PreTrainedTokenizerBase
import vllm.envs as envs
from vllm.engine.protocol import EngineClient
from vllm.entrypoints.generate.base.serving import GenerateBaseServing
from vllm.entrypoints.openai.engine.protocol import (
DeltaMessage,
ErrorResponse,
RequestResponseMetadata,
UsageInfo,
)
from vllm.entrypoints.openai.engine.serving import OpenAIServing
from vllm.entrypoints.openai.models.serving import OpenAIServingModels
from vllm.entrypoints.serve.engine.typing import SpeechToTextRequest
from vllm.entrypoints.serve.utils.api_utils import get_max_tokens
@@ -87,7 +87,7 @@ def asr_inter_chunk_separator(
return "" if language and language.lower() in no_space_languages else " "
class OpenAISpeechToText(OpenAIServing):
class SpeechToTextBaseServing(GenerateBaseServing):
"""Base class for speech-to-text operations like transcription and
translation."""
@@ -475,7 +475,7 @@ class OpenAISpeechToText(OpenAIServing):
list_result_generator: list[AsyncGenerator[RequestOutput, None]] | None = None
input_len = (
OpenAISpeechToText._get_decoder_prompt_len(engine_inputs)
SpeechToTextBaseServing._get_decoder_prompt_len(engine_inputs)
if request.use_beam_search
else 0
)