Vincent Lostanlen

Results 194 comments of Vincent Lostanlen

Hello, unfortunately i do not know the internals of `torch.save` so i'm afraid i can't help you directly. But always keen to review pull requests

I love every item on this list, particularly the Shepard glissando. motion to also credit Risset for this glissando if possible (Shepard made the discrete scale, Risset made the continuous...

Hello @mrava87 , thank you for raising this issue. Indeed it would be good to support NumPy v2 but unfortunately after reading the `utils` code i don't understand if it's...

@janden i see nothing at https://www.kymat.io/_static/ens_logo.jpg at the moment?

> Even if we have averaged (convolved) in frequency and subsampled, don't we record the central frequency here? The default layout is `'joint'` which means that the output is a...

> We do still have a notion of frequency since we perform a convolution along that axis. We then subsample along that axis, but that does not preclude storing those...

Yes, i tend to favor joint formatting. I expect joint formatting to allocate less memory (no need to copy 3D arrays into 2D arrays) and thus a bit faster. It's...

> Well the counterargument here would be to conform with the output format of Scattering1D. The fact that so many of our metadata becomes irrelevant suggests that format='array' would be...

OK! i'm sold on `format='time'` by default Thanks for these benchmarks

Hello both, Paul Smith's RTW department like this. Also I appreciate the repertoire of jazz tunes you have chosen. Congratulations @bmcfee on getting all these examples ready for discussion I...