Ge Zhu (朱舸) comments

Results 6 comments of


                                            Ge Zhu (朱舸)

Code for evaluation in paper

You can try this:https://github.com/gzhu06/Unconditional-Audio-Generation-Benchmark

Question on hilbert transform

Yes, I'd like to. I'll submit a fix some time this or next week. Just want to make sure that to fix this: in 1.7, pytorch uses additional dimension to...

Cannot reproduce cumulative frequency response of SincNet on Speaker-id

I found the similar problem in my applications, the learned filters (i.e. low_hz_, band_hz_) barely changed with epochs. It's basically a mel filter bank

NaN after trained rbig transform

You can download this toy.py.zip to see the working code. (just updated) [toy.py.zip](https://github.com/IPL-UV/rbig/files/6077345/toy.py.zip) thank you so much!

Training Hardware and Issue

Hi, @ayush055 , I also ran into the slow loading issue caused by `musdb` dataloader, so I took @faroit's advice and switch to `trackfolder_fix` dataloader instead. And I got the...

Training Hardware and Issue

Yes, I think the dataloading pipeline of musdb is problematic, `trackfolder_fix` dataloader uses `torchaudio`. I didn't look into the details but I guess there's some bugs in using audio backend...