search

chomeyama / DualCycleGAN

Official implementation of DualCycleGAN for nonparallel audio super resolution
Apache License 2.0
50 stars 5 forks source link

NotImplementedError: Only 2D, 3D, 4D, 5D padding with non-constant padding are supported for now #3

Open harshvardhan-truefan opened 1 year ago

harshvardhan-truefan commented 1 year ago

Hi, I'm trying to train the model on my custom datasets. I have a low resolution dataset of 16000Hz sample rate and high resolution dataset having a sample of 48000Hz. I'm getting the following error while training the model: NotImplementedError: Only 2D, 3D, 4D, 5D padding with non-constant padding are supported for now

This error is occurring during adding padding to the input tensor.

I have attached a screenshot of the error as well. Any idea as to why I'm getting this error ?

Regards, Harsh Screenshot from 2023-08-09 11-48-14

chomeyama commented 1 year ago

Do you have your own special dataset and loader? If you do, please ensure that the dimensions of your input data match the shapes supported, as mentioned in the lines https://github.com/chomeyama/DualCycleGAN/blob/main/dual_cyclegan/model.py#L121 and https://github.com/chomeyama/DualCycleGAN/blob/main/dual_cyclegan/module.py#L134. If the dimensions of your input data are different, you might need to adjust them by reshaping or preprocessing to make them fit the supported shapes.