Joseph Turian

I have a similar issue on a lambdalabs gpu.1x.a6000 Here's what I do in the shell: ``` sudo apt-get update #sudo apt-get -y upgrade sudo -H pip3 install --upgrade pip...

@Gad1001 ``` $ pip install torch==1.9.0+cu111 torchvision==0.10.0+cu111 torchaudio==0.9.0 -f Usage: pip install [options] [package-index-options] ... pip install [options] -r [package-index-options] ... pip install [options] [-e] ... pip install [options] [-e]...

@Ease78 why don't you try it? I would use it.

scipy is generally fine. Just librosa with it's weird numba pin sucks so much.

I am happy the beta test when the branch is ready!

``` pip3.8 install --upgrade git+https://github.com/csteinmetz1/auraloss.git@librosa ``` appears to work. If you submit a PR, I can also code review the diff if you like

For example `torch.cdist`. And I find myself needing this again :)

Or a workaround that if I pass a batch of audio in and a batch of audio out, that I get an in x out return tensor.

Right. The main gnarly bit is if you have a multiscale STFT representation, how to correctly scale each STFT FFT size so that the scores are identical to the existing...

Related https://github.com/turian/torchsynth/issues/244