Sevag H

In hindsight I'm kinda uncertain where the "disable_cupy", "clear_cupy_cache", etc. functions would go, or if they even belong in the package.

If you have access to the MATLAB Wavelet Toolbox (for the CQT/ICQT functions: https://www.mathworks.com/help/wavelet/ref/cqt.html), I have written this algorithm for Harmonic/Percussive/Vocal source separation (loosely based on an iterative version of...

Yes, it's inspired by his algorithm (2-pass with large window + small window), and also Fitzgerald later revisited and added a multipass version with CQT for voice separation: https://arrow.tudublin.ie/cgi/viewcontent.cgi?article=1007&context=argart

Also there could be a "super-performant" config with cupy, stacking multiple 1D FFTs (respecting GPU memory allocation limits), and using pinned host/gpu memory and FFT plans - I'll continue working...

Optimized every slow line (discovered through kernprof + line_profiler): https://github.com/sigsep/sigsep-mus-eval/compare/master...sevagh:feat/cupy-accel This leads to just about 1 minute to compute the IRM mask and perform a BSS evaluation on 1 full-length...

How can I run the tests? `python setup.py test`?

OK. My most recent commits get the regression tests passing. Casting explicitly to float32 was creating huge errors in SAR/SIR/ISR, so I just removed them. I made the cupy install...

As for potential code changes that can deal with such high traffic - maybe multithreading, or sendmmsg (https://github.com/sleinen/samplicator/issues/62)?

@jonahbron works like a charm, thanks: ``` sevagh:repos $ rustup run avr rustc --version rustc 1.37.0-dev ```

There is an issue on the previous version of the repo: https://github.com/hishamhm/htop/issues/899 Implementing generic custom meters should solve the problem of vendor neutrality.