[Frontend] add 'verbose_json' and 'timestamp' feature on Whisper Transcription/Translation (#24209)

Signed-off-by: sangbumlikeagod <oironese@naver.com>
Signed-off-by: sangbumlikeagod <98077576+sangbumlikeagod@users.noreply.github.com>

vllm/model_executor/models/interfaces.py

@@ -837,6 +837,10 @@ class SupportsTranscription(Protocol):
     Transcription models can opt out of text generation by setting this to
     `True`.
     """
+    supports_segment_timestamp: ClassVar[bool] = False
+    """
+    Enables the segment timestamp option for supported models by setting this to `True`.
+    """
 
     def __init_subclass__(cls, **kwargs):
         super().__init_subclass__(**kwargs)

vllm/model_executor/models/whisper.py

@@ -791,6 +791,7 @@ class WhisperForConditionalGeneration(
 
     # Whisper only supports audio-conditioned generation.
     supports_transcription_only = True
+    supports_segment_timestamp = True
     supported_languages = ISO639_1_SUPPORTED_LANGS
 
     @classmethod