[Bugfix][Frontend] Return transcription and translation verbose as float (#49073)
Signed-off-by: Lucas Christian <lucaschgf7@gmail.com>
christian committed
Jul 29, 2026 at 10:16 UTC
e0cfa52d22a00f755a53bbed925b0c92d4290f39
3 files changed
+4
-4
vllm/entrypoints/speech_to_text/base/serving.py
+2
-2
@@ -644,7 +644,7 @@ class SpeechToTextBaseServing(GenerateBaseServing):
644
TranscriptionResponseVerbose(
645
text=text,
646
language=request.language,
647
- duration=str(duration_s),
647
+ duration=duration_s,
648
segments=total_segments,
649
),
650
)
@@ -658,7 +658,7 @@ class SpeechToTextBaseServing(GenerateBaseServing):
658
TranslationResponseVerbose(
659
text=text,
660
language=request.language,
661
- duration=str(duration_s),
661
+ duration=duration_s,
662
segments=total_segments,
663
),
664
)
vllm/entrypoints/speech_to_text/transcription/protocol.py
+1
-1
@@ -384,7 +384,7 @@ class TranscriptionSegment(OpenAIBaseModel):
384
385
386
class TranscriptionResponseVerbose(OpenAIBaseModel):
387
- duration: str
387
+ duration: float
388
"""The duration of the input audio."""
389
390
language: str
vllm/entrypoints/speech_to_text/translation/protocol.py
+1
-1
@@ -357,7 +357,7 @@ class TranslationSegment(OpenAIBaseModel):
357
358
359
class TranslationResponseVerbose(OpenAIBaseModel):
360
- duration: str
360
+ duration: float
361
"""The duration of the input audio."""
362
363
language: str