Shape Mismatch in train_autoencoder.py

[shwj114514]

I encountered an issue where the WAV file, after passing through the encoder and decoder, does not match the original WAV file in
https://github.com/NilsDem/control-transfer-diffusion/blob/main/train_autoencoder.py#L90

AssertionError: Shape mismatch: x.shape = torch.Size([1, 1, 4096]), y.shape = torch.Size([1, 1, 4608])

I suspect the issue might be related to the version of SciPy. I tried both SciPy 1.7.3(python == 3.12.0) and 1.14.0 (python ==3.10.0), but the problem persists.

[NilsDem]

Sorry for the (very) late response. The issue is probably related with the version cached_conv. I just modified the code and requirements accordingly.