Hi thanks for the great library for textless NLP research!
We are using the hubert-base K = 500 for quantization. But the Tacotron2 Vocoder for hubert-base K = 500 is missing in the checkpoint list (now only up to K = 200 is provided), so we can't resynthesize back to speech signal. Is there a plan to release that model? Thanks in advance!
Hi thanks for the great library for textless NLP research!
We are using the hubert-base K = 500 for quantization. But the Tacotron2 Vocoder for hubert-base K = 500 is missing in the checkpoint list (now only up to K = 200 is provided), so we can't resynthesize back to speech signal. Is there a plan to release that model? Thanks in advance!