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

[Voxtral] Fix speech transcription api (#31388)

Browse Source 
Signed-off-by: Patrick von Platen <patrick.v.platen@gmail.com>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Signed-off-by: bk-201 <joy25810@foxmail.com>
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
Signed-off-by: prashanth058 <prashanth.dannamaneni@uipath.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Jee Jee Li <pandaleefree@gmail.com>
Co-authored-by: bk-201 <joy25810@foxmail.com>
Co-authored-by: prashanth058 <prashanth.dannamaneni@uipath.com>
Co-authored-by: Anexdeus <5142168@mail.ru>
Co-authored-by: Julien Denize <40604584+juliendenize@users.noreply.github.com>
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>
Co-authored-by: Nicolò Lucchesi <nicolo.lucchesi@gmail.com>
This commit is contained in:
Patrick von Platen
2026-01-08 12:34:19 +02:00
committed by GitHub
parent 2972a05473
commit 18d4e481d0
5 changed files with 114 additions and 27 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
vllm/model_executor/models/voxtral.py
View File

@@ -17,7 +17,11 @@ from mistral_common.protocol.instruct.chunk import AudioChunk, RawAudio, TextChu
from mistral_common.protocol.instruct.messages import UserMessage
from mistral_common.protocol.instruct.request import ChatCompletionRequest
from mistral_common.protocol.transcription.request import TranscriptionRequest
from mistral_common.tokens.tokenizers.audio import Audio, AudioEncoder
from mistral_common.tokens.tokenizers.audio import (
Audio,
AudioEncoder,
TranscriptionFormat,
)
from transformers import BatchFeature, TensorType, WhisperConfig
from transformers.tokenization_utils_base import TextInput
@@ -157,13 +161,17 @@ class VoxtralProcessorAdapter:
# pad if necessary
# TODO(Patrick) - remove once mistral-common is bumped
sig = inspect.signature(self._audio_processor.pad)
if "is_online_streaming" in sig.parameters:
audio = self._audio_processor.pad(
audio, self.sampling_rate, is_online_streaming=False
)
else:
audio = self._audio_processor.pad(audio, self.sampling_rate)
if (
self._audio_processor.audio_config.transcription_format
!= TranscriptionFormat.STREAMING
):
sig = inspect.signature(self._audio_processor.pad)
if "is_online_streaming" in sig.parameters:
audio = self._audio_processor.pad(
audio, self.sampling_rate, is_online_streaming=False
)
else:
audio = self._audio_processor.pad(audio, self.sampling_rate)
audio_tokens = [self.begin_audio_token_id] + [
self.audio_token_id