Sevag H

If I had to guess, I'm not sure what symbol "♯" is, but I use "#", the regular hash mark: https://github.com/sevagh/chord-detection/blob/master/chord_detection/chromagram.py#L8

I don't know which part of your code or operating system is replacing the regular # with a stylish ♯ but the second one is not one of the keys...

> With the federation system we have in place now, we have trained users that want metrics to be preserved (federated) have to use a specific recording rule style name...

Perhaps something changed in numpy since the last time the code was written?

https://github.com/danilobellini/audiolazy/blob/master/audiolazy/lazy_analysis.py#L789 If I add `.take()` here, the array passed to numpy is correct and the thing works: ``` # Normalization to the [-1; 1] range if normalize: steps = Stream(wnd).blocks(hop).map(np.array).take()...

I'm only running it on one pair of clips - batch size makes sense but I don't know how to set no_grad and batch_sizes with `cdpam`. I'm following the usage...

If I disable GPU and run with `torch.no_grad()`, it seems to be working (however, extremely slowly - after 30 min I was too impatient). The files I'm comparing are 8...

That sounds good. Given PostgREST + NGINX is the recommended/common deployment, one could easily expose the internal metrics prefix at their desired location.

Why would I not just do (pseudocode): ``` band_gains = float[24]; for (j = 0; j < nfft; ++j) float frequency_bin = j * sample_rate/nfft; if (band_0_left < frequency_bin <...

Where do these magic values come from? ``` static const opus_int16 eband5ms[] = { /*0 200 400 600 800 1k 1.2 1.4 1.6 2k 2.4 2.8 3.2 4k 4.8 5.6...