-
Notifications
You must be signed in to change notification settings - Fork 1
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.
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
License #1
Comments
Hi, thanks for your interest. |
Hi, It's the most clean one I've seen, following PyTorch semantics and so on. I was thinking that it would be good to generate a functional from this function and a nn.Module registering the kernel of the convolution as a buffer (not to be recomputed each time you apply an istft) I have also seen the fft version but I don't think it were faster than convolutional version (did you compared it?). Anyway, even if the code comes from keunwoochoi, I saw you've applied several changes and copyright "only" covers the usage of exactly the same code. As you rewrote his code it belongs to you in practice. Edit: |
Wow, thank you for saying so.
-> Yes, I've already done that privately and works seemlessly. The only reason I didn't show that work here is to compare my code's speed performance with keunwoochoi's implementation. As his code doesn't generate nn.Module, I didn't do that here.
-> Yes, I've experimented that in inspection.ipynb, and found that my implementation is 2~5x slower than keunwoochoi's. If speed matters to you, you'd better use
-> I'm glad to hear that. I'll be right back with commit that adds license. |
I've added a license. da9a824 |
Hi, just to be sure, have you considered at the time of timing that cuda is asynchronous? |
Seems like GitHub can't load IPython notebook these days. :( I just did If you create a Pull Request that adds code for fair comparison or precision improvement, I would love to merge it! |
Hi,
Do you plan to add open-source license? It would be super cool.
The text was updated successfully, but these errors were encountered: