Sub-band model

[HWLhsu]

Hi, thanks for sharing this excellent project.
I took a great interest in the model(Delayed Sub-Band LSTM) posted in the paper.
I've tried my hard to reproduce this model, but still can't get a well performance.
Can you release the code about your sub-band models ?
Thanks a lot!

[haoxiangsnr]

Hi, thanks for your attention. I haven't enough time to update the code of the delayed sub-band LSTM recently.

But you could attempt to delete the code corresponding to the full-band model in the FullSubNet project retaining the code about the cIRM and the training processing. Then you will get the delayed sub-band LSTM.

[HWLhsu]

OK, I'll try it.
Thanks for your advice.