fixed the _inference_sample method to use the correct configuration p…

[AdrLfv]

I made two main modifications to the code to resolve the error that was occurring when running the SepReformer model:

1. Write Permission Issue:

• The initial problem was that the code was trying to write output files to the same directory as the input file (/home/user/programs/SepReformer/folder/), but this directory had read-only permissions (dr-x------).

I modified the code to create a separate output directory (output) in the model folder where we have write permissions.

2. Configuration Path Correction:

After the first fix, a KeyError: 'data' error appeared because the configuration keys being used were incorrect.
I fixed the configuration access by using the correct structure found in the configs.yaml file:
self.fs = self.config['dataset']['sampling_rate'] instead of self.config['data']['fs']
self.stride = self.config['model']['module_audio_enc']['stride'] instead of self.config['engine']['stride']

I removed the reference to self.frame_size which was unnecessary