hubert

hubert copied to clipboard

Hi, thank you for your great work, it helps me a lot! I have a question about the Hubert discrete model. Have you done any kind of training for this...

slimsushi

where are the discrete units weights from?

1

in the train.py file, you have an argument named `--warmstart` to allow "initialize[ing] from the fairseq HuBERT checkpoint". I wonder which checkpoint is it since fairseq offers a lot of...

thivux

RuntimeError: stack expects each tensor to be equal size

3

Hello there. I'm in the process of setting up a Colab notebook to train a couple models needed for inference in Soft-VC, mostly for personal ease-of-access. One of the first...

rkjoan

Creating length.json

2

how to properly create length.json file?

danmathewsrobin

Can you provide the training script to build the HuBERT-Discrete Content Encoder? Thanks!

1

Dear, thanks for your great work, can you point me how to rebuild HuBert-Discrete model?

kelvinqin

why 16kHz ? why not 22kHz ?

1

what is the reason? the quality should be better...

salisbury-espinosa

Hi, thanks for sharing this great work. What should be the validation/training accuracy/loss to ensure a good enough training ?

EmreOzkose

how to train hubert-soft content encoder using Mandarin Chinese dataset?

1

my problem is as title, how to train hubert-soft content encoder using Mandarin Chinese dataset? still using the pretrained hubert model to extract Discrete Speech Units ? (encode dataset step)...

joan126

(officially released model:https://github.com/facebookresearch/fairseq/tree/main/examples/hubert) What is the difference between the model you released and the officially released model? Or how is the model you released trained?

wyp19930313

yaoao2017