Zhaoci Liu
Results
3
comments of
Zhaoci Liu
There is an example of reconstruction in the [tutorial](https://pycwt.readthedocs.io/en/latest/tutorial.html), and after looking at the output [picture](https://pycwt.readthedocs.io/en/latest/) it is true that it is not fully reconstructed. But after fixing the bug...
Using `repeat_interleave` to replace `for` is a good idea. But could `output.reshape(N,-1,D)` get the right result since different samples in a batch have different lengths? Maybe replacing `for` with `repeat_interleave`...
The problem was solved by using soundfile to save audio Modify the file `preprocessor/ljspeech.py`: ```python # from scipy.io import wavfile import soundfile as sf # ... wav = wav /...