Deep Learning for Population-Dependent Controls in Mean Field Control Problems