Sam Lapp

Loading fragments of file is slow in newer version

1

**Describe the bug** Loading a piece of a longer audio file is much slower in v0.9.2 than it was in v0.8.1. For instance, loading a 2-hour wav file takes 440...

Opening file in write mode removes existing metadata

2

It seems that simply opening a SoundFile (version 0.10.3.post1, which is what I get when I run `pip install soundfile`) in write mode removes existing metadata in the wav file....

If I create ScoreCAM with a model on a non-cpu device (eg, 'mps' on Apple Silicon), and pass a tensor on the same device, it raises an error at the...

Running `GuidedBackpropReLUModel.__call__(tensor)` twice in a row on the same tensor produces unexpected results because the method sets `tensor`'s requires_grad to True and does not detach afterward. This causes two unexpected...

GuidedBackpropReLUModel always uses a batch size of 1, this is hard-coded into the implementation eg here: https://github.com/jacobgil/pytorch-grad-cam/blob/2183a9cbc1bd5fc1d8e134b4f3318c3b6db5671f/pytorch_grad_cam/guided_backprop.py#L89 Will the Guided Backprop be broken by batching? I could see how batching...

Bug: ValueError when loading model previously cached but now missing

5

### What happened? Loading a tensorflow hub model that I previously loaded results in an error. As described in https://github.com/tensorflow/hub/issues/575, it seems that TF Hub looks for the model in...

Have weighting curves been validated?

6

The README mentions that some parts of this repo have not been validated. This seems to be the best Python resource for A, C weighting of digital audio, and it...

sf_set_string raises an error if passed an empty string

1

I would expect `sf_set_string` to be able to set a file's metadata field to an empty string (`''`). https://github.com/libsndfile/libsndfile/blob/7f7c5b0bc9012d1d4528eefe9b8e5f1b3ee2f916/src/sndfile.c#L1635 However, passing `''` to `sf_set_string()` leads to a "bad string" error,...

does the TFLite model allow batching?

1

I am using the TF Lite checkpoint https://github.com/kahst/BirdNET-Analyzer/raw/main/checkpoints/V2.4/BirdNET_GLOBAL_6K_V2.4_Model_FP16.tflite it seems I cannot pass batched samples, as passing any np array shape other than [1,144000] results in the model complaining the...

I get tons of messages like ``` [src/libmpg123/layer3.c:INT123_do_layer3():1801] error: dequantization failed! Warning: Xing stream size off by more than 1%, fuzzy seeking may be even more fuzzy than by design!...