Can deep learning provide a generalizable model for dynamic sound encoding in auditory cortex?