error while trying to train about svoice HOT 5 CLOSED

Hi @th3geek,
From the comment it seems like there is a mismatch between the sizes of the estimate and target. How many sources do you want to separate to? The default is 2.

from svoice.

Hi,

Yes, I'm trying to separate 4 sources. I've looked through the config and don't see the option for specifying the number of sources?

from svoice.

You are right, I should have been clearer about that.
You should set swave.C=4. See here: https://github.com/facebookresearch/svoice/blob/master/conf/config.yaml#L66

from svoice.

That worked! Thank you.

from svoice.

Hi, @adiyoss @th3geek as mentioned in the paper this model can separate an unknown number of sources in overlapped speech.
Can you tell how we can do this example:- mix files have a different number of source's mixed audio than what will be the value of swave.C or there is a different approach to solve such problem

from svoice.