alec-wright / automated-guitarampmodelling Goto Github PK

hey, Alec
as it stands, the project is in the open but distributed without a open source license
would you consider licensing it accordingly?

you can learn more here: https://choosealicense.com/licenses/
thanks

process at time of error
$ python dist_model_recnet.py -l "RNN3-acoustic1-pre"
no saved model found, creating new network
cuda device not available/not selected
unimplemented audio data type conversion...
unimplemented audio data type conversion...
unimplemented audio data type conversion...
unimplemented audio data type conversion...
unimplemented audio data type conversion...
unimplemented audio data type conversion...
Epoch: 1
current learning rate: 0.005
Epoch: 2

Error
Traceback (most recent call last):
File "dist_model_recnet.py", line 190, in
val_output, val_loss = network.process_data(dataset.subsets['val'].data['input'][0],
File "C:\Users\nickg\Automated-GuitarAmpModelling\CoreAudioML\networks.py", line 115, in process_data
output[l * chunk:(l + 1) * chunk] = self(input_data[l * chunk:(l + 1) * chunk])
RuntimeError: The expanded size of the tensor (100000) must match the existing size (20733) at non-singleton dimension 0. Target sizes: [100000, 1, 1]. Tensor sizes: [20733, 1, 1]

System
python 3.8.6
pytorch==0.1.2
torchvision==0.1.6
pytorch_lightning==1.1.0
numpy==1.19.4
scipy==1.5.4
matplotlib==3.3.3

Hello,

I'm using wav files with 44.1k sample rate exported as 32 bit float for sample data.
when I run this training, I get the following warning:

unimplemented audio data type conversion...

And when I run a prediction using the output model, it seems there is no difference between the original signal and the processed signal.

I also tried with some of the sample data provided in this repo, and that seems to work fine... can you please detail what kind of file format it is expecting ? sample rate, bit format, length / duration, etc...

Thanks !
Stephane