I'm trying to use Facebook's SVoice https://github.com/facebookresearch/svoice to separate the sources. However, while I got the training to work, I am running into issues when I try to use the evaluate function. I followed using the commands as instructed:
python -m svoice.evaluate /home/username/_SVoice2/outputs/exp_/model.pt /home/username/_SVoice2/egs/jsonpath_with_mixjson_and_all_target_separated_channels_jsonfiles
However, I got this error:
Traceback (most recent call last):
File "/home/username/anaconda3/envs/svoice3/lib/python3.7/runpy.py", line 193, in _run_module_as_main
"__main__", mod_spec)
File "/home/username/anaconda3/envs/svoice3/lib/python3.7/runpy.py", line 85, in _run_code
exec(code, run_globals)
File "/home/username/_SVoice2/svoice/evaluate.py", line 215, in <module>
main()
File "/home/username/_SVoice2/svoice/evaluate.py", line 208, in main
sisnr, pesq, stoi = evaluate(args)
File "/home/username/_SVoice2/svoice/evaluate.py", line 60, in evaluate
model = deserialize_model(model)
File "/home/username/_SVoice2/svoice/utils.py", line 35, in deserialize_model
klass = package['class']
TypeError: 'SWave' object is not subscriptable
How can I fix this?