Open zbpjlc opened 4 months ago
SpeechT5 , how can i extract the fixed transorformer layer feature; in the hubert paper: so,in the SpeechT5,how can i do,Is there any relevant code ?
for the second iteration
Sorry, we did not provide the scripts but you can modify from here and get the last layer features by passing "feature_only" parameter during forward which will return the feature in here.
SpeechT5 , how can i extract the fixed transorformer layer feature; in the hubert paper:
so,in the SpeechT5,how can i do,Is there any relevant code ?