How do I use the separation feature?

[zxxwxyyy]

Hi, love the model – super cool stuff! Thank you for open-source the code!!

I was playing around with the notebook demos and couldn't find any demos on the separation feature. Really keen on trying that out. Would you kindly point me in the right direction or share a quick how-to?

Thank you so much!

[Logicino]

Hello, have you solved this problem? I run into the same problem
I tried to import the "MSDMSeparator" like below

from main.separation import MSDMSeparator

num_steps = 20

import soundfile as sf
audio, sr = sf.read("mix.wav")
print(type(audio))
audio = torch.from_numpy(audio.T)

schedule = KarrasSchedule(sigma_min=1e-4, sigma_max=20.0, rho=7)(num_steps, DEVICE)
separator = MSDMSeparator(model=model, stems=[], sigma_schedule=schedule)
separator.separate(audio)

however it comes with problems:

due to the batch_size, _, length_samples = mixture.shape
I don't understand what shape should the mixture be
How did you solve the problem? I really appreciate it! Thank you!