Getting matrix size 0x87 while training

[AsimFayyazRaja]

OS: Ubuntu 16.04 Environment: Local using pip

I have some wav files which are appended together using pydub and then raw file is made from the wav file like this:

sox joinedFile_clean.wav --bits 16 --encoding signed-integer --endian little ../clean.raw
sox joinedFile_noisy.wav --bits 16 --encoding signed-integer --endian little ../noisy.raw

then the raw files are being passed to denoise_training like this:

./denoise_training clean.raw noisy.raw count > training.f32

but the output I am getting is:

matrix size: 0 x 87

Can anyone tell me why is this issue happening? Am I converting it correctly to raw? The inference is working fine.

[jeremy110]

Your count should be set 50000 or 500000.The problem can be solved.

[Mikelew88]

Having the same issue, has anybody solved it?

[Mikelew88]

I now understand. For anyone else, you must specify and integer for count, Jeremy suggests 50000 or 500000, in the ./denoise_training command. So for example:

./denoise_training clean.raw noisy.raw 50000 > training.f32

[pengzhendong]

You must set the maxCount refers to src/denoise.c. The code will generate maxCount (default: 0) frames in loop.