the results and training loss

Question

the results and training loss

DingGitTemp opened this issue 5 months ago · comments

I attempted to train and test the model on the voicebank-demand dataset, but the results were not satisfactory. The enhanced speech couldn't be recognized as human voice . Are there any parameters that need to be reset?Additionally, during the training process, the loss of the training set consistently remained around 700. Is this normal?

Simon Welker · Answer 1 · Fri Mar 08 2024 19:27:55 GMT+0800 (China Standard Time)

Hi, did you downsample your version of the VB-DMD dataset to 16 kHz? The model is by default designed for 16 kHz.

DingGitTemp · Answer 2 · Fri Mar 08 2024 21:43:04 GMT+0800 (China Standard Time)

The issue has been resolved; it turns out I hadn't downsampled the data to 16K. Thank you very much for your response; this truly is a remarkable piece of work.

Simon Welker · Answer 3 · Fri Mar 08 2024 21:59:27 GMT+0800 (China Standard Time)

Thanks! Happy to hear that it works now.