GuitarML/PedalNetRT

Python error when training

jackcolt opened this issue · 5 comments

jackcolt commented

Really cool stuff here. Thank you.

When running the train.py, I get the following cryptic python error. Any thoughts? I can train the LTSM models no problem using the source straight out of the repo.

Traceback (most recent call last):
File "train.py", line 56, in
main(args)
File "train.py", line 21, in main
model = PedalNet(vars(args))
File "/home/jack/PedalNetRT/model.py", line 118, in init
self.hparams = hparams
File "/home/jack/.local/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1225, in setattr
object.setattr(self, name, value)
AttributeError: can't set attribute

mishushakov commented

hey, what command line arguments are you using with train.py?

jackcolt commented

Thanks for responding. I'm just calling with the params from the readme here in the repo.

python train.py data/ts9_test1_in_FP32.wav data/ts9_test1_out_FP32.wav

GuitarML commented

@jackcolt Haven't been able to reproduce what you're seeing, are you still getting this or has it been resolved?

dobell-christopher commented

Change line 118 in module.py from
self.hparams = hparams
to:
self.hparams.update(hparams)

You must have a newer version of pytorch as self.hparams = hparams no longer works in the new version

  • 👍1
JamesJieranShen commented

For proper model restoration, it also seems necessary to call self.save_hyperparameters() afterwards.