Non-zero exit status

[Rein Liu]

Hi Marco! Thanks for this interesting research and repo. I'm trying to run your codes based on some of SFX datasets. However, I'm having some issues with subprocess:

drumgan_evaluation/shell_scripts/fad.sh: 3: Syntax error: "(" unexpected
Traceback (most recent call last):
File "train_ccwavegan_cchifigan.py", line 249, in
device=device
File "train_ccwavegan_cchifigan.py", line 219, in train_model
verbose=verbose
File "/home/rein/OneDrive/GitHub/hifi-wavegan-footsteps/architectures/ccwavegan_cchifigan.py", line 266, in train
verbose=verbose,
File "/home/rein/OneDrive/GitHub/hifi-wavegan-footsteps/utils/utils_training.py", line 205, in compute_fad_at_batch
"--output="+output_path
File "/home/rein/anaconda3/lib/python3.7/subprocess.py", line 376, in check_output
**kwargs).stdout
File "/home/rein/anaconda3/lib/python3.7/subprocess.py", line 468, in run
output=stdout, stderr=stderr)
subprocess.CalledProcessError: Command '['sh', 'drumgan_evaluation/shell_scripts/fad.sh', '--real=/home/rein/OneDrive/GitHub/hifi-wavegan-footsteps/checkpoints-footstep/2022-11-12_16h8m/fad/real_audio.csv', '--synth=/home/rein/OneDrive/GitHub/hifi-wavegan-footsteps/checkpoints-footstep/2022-11-12_16h8m/fad/synth_audio.csv', '--output=/home/rein/OneDrive/GitHub/hifi-wavegan-footsteps/checkpoints-footstep/2022-11-12_16h8m/fad']' returned non-zero exit status 2.

I'm actually running the latest pytorch 1.13 and cuda 11.7. I see in your repository it was torch 1.9 but for me the training was running. The only problem I'm encounting is the evaluation with FAD at training. Do you have any insight about this? Thank you in advance.