Inference script to hear samples from AVSpeech Model ?

I have implemented the [AVSpeech](https://github.com/asteroid-team/asteroid/tree/master/egs/avspeech/looking-to-listen
) training and eval. How can hear out the sound and check how the model performs? Also, what is the ideal SNR for the model?

How can I find the code to infer to hear the outputs? The checkpoint is `best.pth`?

Start with egs/avspeech/looking-to-listen and the existing AVSpeech training and evaluation setup. Trace how the best.pth checkpoint is used, then identify the documented path for producing audio outputs and checking SNR; done means a newcomer can run inference and listen to the samples.

Written by the indexing model from the issue text.