Iver Jordal

Like here: https://librosa.org/doc/latest/generated/librosa.feature.melspectrogram.html#librosa.feature.melspectrogram

These examples should make it easier to the user to understand what a transform is doing

Change default noise_rms="relative_to_whole_input" in AddShortNoises

1

The default value for noise_rms is currently "relative", but "relative_to_whole_input" might actually be a better default, given that the loudness of the given signal can vary from very silent to...

Fix "is directory" check in test_yolo.py

1

I'm proposing this change because the existing implementation gave me an error like this: ``` Traceback (most recent call last): File "test_yolo.py", line 194, in _main(parser.parse_args()) File "test_yolo.py", line 119,...

Feature request: Add support for ViSQOL

4

Thanks for making speechmetrics! Here's a candidate implementation of ViSQOL: https://github.com/google/visqol

Fix bug where ISTFT with device="cuda" and move between devices didn't work

1

I've added a test that previously failed*, but passes with the few patches I applied \* `E RuntimeError: Input type (torch.cuda.FloatTensor) and weight type (torch.FloatTensor) should be the same`

Explain that AreYouSure does not monitor input elements without the `name` attribute