Batch load for pywt.cwt

[SongRb]

I'm working on a machine learning application and have to do transformation on large scale of data. I want to call the function like pywt.cwt((batch_size, signal_length)), but it seems only dim=1 supported. Is there any plan for supporting this feature?

[grlee77]

Hi @SongRb:

An implementation of this has been proposed in #509.

Any feedback there would be welcome. In the case of data with shape (batch_size, signal_length), you could leave the default axis=-1 argument since the signal is along the last axis of the input. The output would have shape (len(scales), batch_size, signal_length).

Also, another recent PR allows setting mode='fft' to do FFT-based convolutions for the CWT. This should have a performance benefit for larger data / scales.