[Audio] Bundle get_generation_prompt() params into SpeechToTextParams (#36268)

Signed-off-by: Ekagra Ranjan <3116519+ekagra-ranjan@users.noreply.github.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:
Ekagra Ranjan
2026-04-22 12:24:18 +08:00
committed by GitHub
co-authored by gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> Cyrus Leung
parent 6d09769700
commit 4eafc72928
18 changed files with 214 additions and 161 deletions
@@ -4,7 +4,6 @@
import asyncio
import math
from collections.abc import AsyncGenerator, Iterable, Iterator, Mapping
from typing import Literal
import numpy as np
import torch
@@ -18,6 +17,7 @@ from mistral_common.tokens.tokenizers.audio import AudioConfig
from vllm.compilation.decorators import support_torch_compile
from vllm.config import ModelConfig, SpeechToTextConfig, VllmConfig
from vllm.config.speech_to_text import SpeechToTextParams
from vllm.engine.protocol import StreamingInput
from vllm.envs import VLLM_ENGINE_ITERATION_TIMEOUT_S
from vllm.inputs import PromptType, TokensPrompt
@@ -465,14 +465,13 @@ class VoxtralRealtimeGeneration(VoxtralForConditionalGeneration, SupportsRealtim
# for speech-to-text transcription
def get_generation_prompt(
cls,
audio: np.ndarray,
model_config: ModelConfig,
stt_config: SpeechToTextConfig,
language: str | None,
task_type: Literal["transcribe", "translate"],
request_prompt: str,
to_language: str | None,
stt_params: SpeechToTextParams,
) -> PromptType:
audio = stt_params.audio
model_config = stt_params.model_config
stt_config = stt_params.stt_config
language = stt_params.language
tokenizer = cached_tokenizer_from_config(model_config)
audio = Audio(audio, int(stt_config.sample_rate), format="wav") # lossless