|
|
|
@ -1191,8 +1191,9 @@ class DecodingTask:
|
|
|
|
|
DecodingResult(
|
|
|
|
|
audio_features=features,
|
|
|
|
|
language=language,
|
|
|
|
|
language_probs=probs) for features, language, probs in
|
|
|
|
|
zip(audio_features, languages, language_probs)
|
|
|
|
|
language_probs=probs)
|
|
|
|
|
for features, language, probs in zip(audio_features, languages,
|
|
|
|
|
language_probs)
|
|
|
|
|
]
|
|
|
|
|
|
|
|
|
|
# repeat the audio & text tensors by the group size, for beam search or best-of-n sampling
|
|
|
|
|