sreyas-mohan / deepfreq Goto Github PK

Hi,
upon running the ´example_notebook.ipynb´ I discovered a dimensionality bug.
In code cell 7

with torch.no_grad():
    fr_0dB = fr_module(torch.tensor(signal_0dB[idx][None]))
    nestimate_0dB = fc_module(fr_0dB).numpy().round()
    fr_10dB = fr_module(torch.tensor(signal_10dB[idx][None]))
    nestimate_10dB = fc_module(fr_10dB).numpy().round()
    fr_50dB = fr_module(torch.tensor(signal_50dB[idx][None]))
    nestimate_50dB = fc_module(fr_50dB).numpy().round()
    
fr_0dB = fr_0dB.numpy()    
fr_10dB = fr_10dB.numpy()    
fr_50dB = fr_50dB.numpy()

the following Runtime error occurs, which points to a dimensionality problem IMHO:

RuntimeError: size mismatch, m1: [1 x 308], m2: [200 x 1] at /pytorch/aten/src/TH/generic/THTensorMath.cpp:41

My environment:

• numpy - version 1.18.3
• torch - version 1.5.1

As I was running the notebook to better understand the paper, I am not sure how to resolve this mismatch. Can you let me know how to fix it?
Big thank you for posting your code, btw. I really appreciate the effort.
Looking forward to hear from you!

Excuse me, if I want to reduce the frequency training range of the model from -0.5 to -0.1 to 0.1, so that the number of output points remains unchanged and the frequency output range decreases, but the frequency resolution becomes higher, does this leave a parameter interface in your parameters? Or what can be improved to achieve,thank you!

When I run this program, an error appears: UserWarning: Using a target size (torch.Size([256, 1000])) that is different to the input size (torch.Size([256, 1320])). This will likely lead to incorrect results due to broadcasting. Please ensure they have the same size.
Looking forward to your reply