Remove rebuttal comments

main.py

@@ -227,11 +227,9 @@ def random_resize_collate(batch, default_collate=torch.utils.data.dataloader.def
     assert args.n_bits == 8
     if args.binarize:
         rescaling = lambda x : (binarize_torch(x) - .5) * 2.  # binarze and rescale [0, 1] images into [-1, 1] range
-        #rescaling = lambda x : binarize_torch(x).float()  # binarze [0, 1] images NOTE: temporary for rebuttal
     else:
         rescaling = lambda x : (x - .5) * 2.  # rescale [0, 1] images into [-1, 1] range
     rescaling_inv = lambda x : .5 * x + .5
-    #rescaling_inv = lambda x : x  # NOTE: temporary for rebuttal
     ds_transforms = transforms.Compose([transforms.ToTensor(), rescaling])
 
     train_loader = torch.utils.data.DataLoader(datasets.MNIST(args.data_dir, download=True,
@@ -278,12 +276,6 @@ def rescaling(x):
             x = (2. / (n_bins - 1)) * x - 1.
             return x
         rescaling_inv = lambda x : .5 * x + .5  # rescale [-1, 1] range to [0, 1] range
-        #def rescaling_inv(x):
-        #    # rescale images from [-1, 1] to [0, n_bins-1]
-        #    x = ((n_bins - 1) / 2) * (x + 1.)
-        #    # increase bit depth to [0, 255] range
-        #    x = x * depth_divisor
-        #    return x
 
     # NOTE: Random resizing of images during training is not supported for CelebA-HQ. Will use 256x256 resolution.
     def get_celeba_dataloaders():

utils.py

@@ -336,7 +336,6 @@ def discretized_mix_logistic_loss_1d_averaged(x, ls):
 
 def _binarized_label(x):
     assert x.size(1) == 1
-    # NOTE: temporarily disabled for rebuttal, checking [0, 1] images + mask conditioning
     x = x * .5 + .5  # Scale from [-1, 1] to [0, 1] range
     x = binarize_torch(x)  # binarize image. Should be able to just cast,
                             # since x is either 0. or 1., but this could avoid float
@@ -503,7 +502,6 @@ def sample_from_binary_logits(l, coord1, coord2):
     assert l.size(1) == 2
     l = l[:, :, coord1, coord2]
     pixels = torch.distributions.categorical.Categorical(logits=l).sample()
-    # NOTE: Disabled for {0, 1} input rebuttal experiments
     pixels = pixels * 2. - 1.
     return pixels.unsqueeze(1)