remove output scale param from trainings as this is now done by rescaling the precision factor

mu_map/training/cgan.py

@@ -258,12 +258,6 @@ if __name__ == "__main__":
         default="data/initial/",
         help="the directory where the dataset for training is found",
     )
-    parser.add_argument(
-        "--output_scale",
-        type=float,
-        default=1.0,
-        help="scale the attenuation map by this coefficient",
-    )
     parser.add_argument(
         "--input_norm",
         type=str,
@@ -398,8 +392,6 @@ if __name__ == "__main__":
     elif args.input_norm == "gaussian":
         transform_normalization = GaussianNormTransform()
 
-    transform_augmentation = ScaleTransform(scale_outputs=args.output_scale)
-
     data_loaders = {}
     for split in ["train", "validation"]:
         dataset = MuMapPatchDataset(
@@ -410,7 +402,6 @@ if __name__ == "__main__":
             shuffle=not args.no_shuffle,
             split_name=split,
             transform_normalization=transform_normalization,
-            transform_augmentation=transform_augmentation,
             logger=logger,
         )
         data_loader = torch.utils.data.DataLoader(

mu_map/training/default.py

@@ -151,12 +151,6 @@ if __name__ == "__main__":
         default="data/initial/",
         help="the directory where the dataset for training is found",
     )
-    parser.add_argument(
-        "--output_scale",
-        type=float,
-        default=1.0,
-        help="scale the attenuation map by this coefficient",
-    )
     parser.add_argument(
         "--input_norm",
         type=str,
@@ -288,8 +282,6 @@ if __name__ == "__main__":
     elif args.input_norm == "gaussian":
         transform_normalization = GaussianNormTransform()
 
-    transform_augmentation = ScaleTransform(scale_outputs=args.output_scale)
-
     data_loaders = {}
     for split in ["train", "validation"]:
         dataset = MuMapPatchDataset(
@@ -300,7 +292,6 @@ if __name__ == "__main__":
             shuffle=not args.no_shuffle,
             split_name=split,
             transform_normalization=transform_normalization,
-            transform_augmentation=transform_augmentation,
             logger=logger,
         )
         data_loader = torch.utils.data.DataLoader(