Training code for Full-Attention-Residual-DU-Net

Attention gates placed at two places.

  1. In the skip connection between encoder1 and decoder1
  2. In the skip connection between encoder2 and decoder2