Global consistency loss appears like val loss

[GozdeUnver]

Hello,
I went through your code and it looks like in train.py the global consistency is calculated and printed for the validation. Even though this doesn't affect the model training, I mean the global consistency loss is still affecting the model in train mode, it is printed out as validation. Additionally, there is no loop for validation run. :

for j in range(0, len(src_tiles), tile_batch_size):
      src_tiles_ = src_tiles[j: j + tile_batch_size]
      dst_tiles_ = dst_tiles[j: j + tile_batch_size]
      dst_fake_tiles_ = dst_fake_tiles[j: j + tile_batch_size]

      # pass real and fake images
      model.set_input_image({'src_real': src_tiles_, 'dst_real': dst_tiles_, 'dst_fake': dst_fake_tiles_})

      # calculate style and content losses, gradients, update network weights
      losses = model.optimize_parameters_image()
      iter_count += 1

      # free up gpu memory
      delete_tensor_gpu({'src': src_tiles_, 'dst': dst_tiles_, 'dst_fake': dst_fake_tiles_})

      # loss logging
      if iter_count % loss_logging_freq == 0:
          losses = {k: round(v, 4) for k, v in losses.items()}
          print('val losses: ', losses)``