Closed cthoyt closed 8 months ago
I am not convinced the layer we proposed back in 2020 is still relevant for current architectures ( see also https://github.com/v0lta/PyTorch-Wavelet-Toolbox/issues/79 ). The idea of having the 'nn' namespace is nice, but we would need something people can just plug in with good performance, the compressed layer will most likely not do that.
Closes #79