Cyrus Asfa
Cyrus Asfa
The 1D examples are still broken on the website https://www.kymat.io/gallery_1d/plot_filters.html#sphx-glr-gallery-1d-plot-filters-py Did we merge @danedane-haider's PR? perhaps the website isn't rendering from `main`?
[pypi](https://pypi.org/project/kymatio/) is still on v0.2.1. we've now released v0.3 - how do we publish to pypi?
thanks for sharing this. I am trying to train PCEN in a torch layer. However, at unpredictable points during training, the loss collapses to `nan` when this is introduced. I'm...
When loading the shift predictor, its dimensions were not being loaded correctly. It does not match the directions of the latent deformator. For example, with a deformator of 64 dimensions,...
In time-frequency scattering we stack the S1 paths along the 2nd axis frequency: https://github.com/kymatio/kymatio/blob/8f2bbdfdbe974de82caf52b07826859c02d3bd24/kymatio/scattering1d/core/timefrequency_scattering.py#L157 This is correct in torch, since the default value for `axis=2`, however, `axis=1` in `numpy_backend` and...
In the implementation of JTFS, we introduced a new backend primitive named swap_time_frequency. This conveniently tranposes the time and frequency axes such that we can perform scattering along the frequency...
`torch.jit.script` throws a `cannot create weak reference to 'numpy.ufunc' object` error when compiled `LogmelFilterBank`. I have fixed this by changing numpy operations to torch (np.maximum) and storing `self.ref` and `self.amin`...
``` runtimeError: Audio file with offset 1472.0742892449407 and duration 10.0 is empty! ``` I've been getting this error about 1-2 hours into training. It happens for a different audio file...