Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

run lfilter on GPU #2

Open
yoyolicoris opened this issue Jan 28, 2023 · 2 comments
Open

run lfilter on GPU #2

yoyolicoris opened this issue Jan 28, 2023 · 2 comments
Assignees
@yoyolicoris
Labels
enhancement New feature or request

Comments

@yoyolicoris
Copy link
Member

yoyolicoris commented Jan 28, 2023 •
edited
Loading

depends on yoyolicoris/audio#12
@yoyolicoris yoyolicoris added the enhancement New feature or request label Jan 28, 2023
@yoyolicoris yoyolicoris self-assigned this Jan 28, 2023
@csteinmetz1
Copy link

You can also implement a fast IIR filter on GPU by approximating it with an FIR filter and then apply the filter in the frequency domain via FFT. You can see an example here: https://github.com/adobe-research/DeepAFx-ST/blob/49bd0c88f3a0ad8b9c45f88200b4caa74952c43d/deepafx_st/processors/autodiff/signal.py#L158

@yoyolicoris
Copy link
Member Author

Thanks for the suggestion, Chris.
I see where you're talking and it's a reasonable workaround in pytorch, but performing FFT is kind of overkill for low-order IIR in my opinion, especially when the input sequence is long.
I'm looking into custom cuda kernel for IIR and have some preliminary results on torchaudio.
This issue can be solved easily after my PR is merged.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@yoyolicoris yoyolicoris

Labels
enhancement New feature or request
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@csteinmetz1 @yoyolicoris