From d100ab38d36ba38d1ae9c55b1134b877438f6875 Mon Sep 17 00:00:00 2001
From: Geet Sethi <geetsethi05@gmail.com>
Date: Mon, 30 Oct 2023 14:01:19 +0530
Subject: [PATCH 1/2] reformat all files

---
 UniTrain/models/classification.py  | 167 ++++++++++++++++++------
 UniTrain/models/segmentation.py    |  25 ++--
 UniTrain/utils/DCGAN.py            |  31 ++++-
 UniTrain/utils/StyleTransfer.py    | 101 +++++++++------
 UniTrain/utils/StyleTransferVGG.py | 199 ++++++++++++++++-------------
 UniTrain/utils/__init__.py         |   9 +-
 UniTrain/utils/classification.py   | 140 +++++++++++---------
 UniTrain/utils/segmentation.py     |  59 +++++----
 8 files changed, 454 insertions(+), 277 deletions(-)

diff --git a/UniTrain/models/classification.py b/UniTrain/models/classification.py
index b3cf3c7..c38e005 100644
--- a/UniTrain/models/classification.py
+++ b/UniTrain/models/classification.py
@@ -14,11 +14,20 @@ def __init__(self, num_classes):
         self.softmax = F.softmax
         self.num_classes = num_classes
 
-
-    def conv_block(self, xb, inp_filter_size, hidden_filter_size, out_filter_size, pool = False):
-        layers = nn.Sequential(nn.Conv2d(inp_filter_size, hidden_filter_size, padding=0, kernel_size=1), nn.BatchNorm2d(hidden_filter_size), nn.ReLU(inplace=True),
-                  nn.Conv2d(hidden_filter_size, hidden_filter_size, padding=1, kernel_size=3), nn.BatchNorm2d(hidden_filter_size), nn.ReLU(inplace=True),
-                  nn.Conv2d(hidden_filter_size, out_filter_size, padding=0, kernel_size=1), nn.BatchNorm2d(out_filter_size), nn.ReLU(inplace=True))
+    def conv_block(
+        self, xb, inp_filter_size, hidden_filter_size, out_filter_size, pool=False
+    ):
+        layers = nn.Sequential(
+            nn.Conv2d(inp_filter_size, hidden_filter_size, padding=0, kernel_size=1),
+            nn.BatchNorm2d(hidden_filter_size),
+            nn.ReLU(inplace=True),
+            nn.Conv2d(hidden_filter_size, hidden_filter_size, padding=1, kernel_size=3),
+            nn.BatchNorm2d(hidden_filter_size),
+            nn.ReLU(inplace=True),
+            nn.Conv2d(hidden_filter_size, out_filter_size, padding=0, kernel_size=1),
+            nn.BatchNorm2d(out_filter_size),
+            nn.ReLU(inplace=True),
+        )
         layers.to(xb.device)
         return layers(xb)
 
@@ -38,7 +47,7 @@ def forward(self, xb):
         y = self.conv_block(y, 512, 256, 1024)
         for i in range(0, 22):
             y = self.conv_block(y, 1024, 256, 1024) + y
-            i+=1
+            i += 1
 
         y = self.conv_block(y, 1024, 512, 2048)
         y = self.conv_block(y, 2048, 512, 2048) + y
@@ -52,7 +61,7 @@ def forward(self, xb):
         y = linear_layer(y)
 
         return y
-    
+
 
 # Define the ResNet-9 model in a single class
 class ResNet9(nn.Module):
@@ -120,28 +129,32 @@ def forward(self, x):
         x = x.view(x.size(0), -1)
         x = self.fc(x)
 
- #ResNet50 functionality addition       
+
+# ResNet50 functionality addition
 class ResNet9_50(nn.Module):
     def __init__(self, num_classes):
         super(ResNet9_50, self).__init__()
-        
+
         self.resnet9 = ResNet9(num_classes)
         self.resnet50 = models.resnet50(pretrained=True)
 
     def forward(self, x):
-         x = self.resnet50(x)
-#GoogLeNet functionality addition
+        x = self.resnet50(x)
+
+
+# GoogLeNet functionality addition
 import torch
 import torch.nn as nn
 import torchvision.models as models
 
+
 class GoogleNetModel(nn.Module):
     def __init__(self, num_classes):
         super(GoogleNetModel, self).__init()
-        
+
         # Load the pre-trained GoogleNet model
         self.googlenet = models.inception_v3(pretrained=True)
-        
+
         # Modify the classification head to match the number of classes in your dataset
         num_ftrs = self.googlenet.fc.in_features
         self.googlenet.fc = nn.Linear(num_ftrs, num_classes)
@@ -156,10 +169,15 @@ def forward(self, x):
 
 
 # Making a custom transfer learning model
-def create_transfer_learning_model(num_classes, model = torchvision.models.resnet18, feature_extract=True, use_pretrained=True):
+def create_transfer_learning_model(
+    num_classes,
+    model=torchvision.models.resnet18,
+    feature_extract=True,
+    use_pretrained=True,
+):
     """
     Create a transfer learning model with a custom output layer.
-    
+
     Args:
         num_classes (int): Number of classes in the custom output layer.
         model(torchvision.models.<ModelName>): Pre-trained model you want to use.
@@ -171,18 +189,19 @@ def create_transfer_learning_model(num_classes, model = torchvision.models.resne
     """
     # Load a pre-trained model, for example, ResNet-18
     model = model(pretrained=use_pretrained)
-    
+
     # Freeze the pre-trained weights if feature_extract is True
     if feature_extract:
         for param in model.parameters():
             param.requires_grad = False
-    
+
     # Modify the output layer to match the number of classes
     num_features = model.fc.in_features
     model.fc = nn.Linear(num_features, num_classes)
-    
+
     return model
 
+
 # Define the ResNet-18 model in a single class
 class ResNet34(nn.Module):
     def __init__(self, num_classes):
@@ -210,15 +229,31 @@ def make_layer(self, out_channels, num_blocks, stride):
         layers.append(self.build_residual_block(self.in_channels, out_channels, stride))
         self.in_channels = out_channels
         for _ in range(1, num_blocks):
-            layers.append(self.build_residual_block(self.in_channels, out_channels, stride=1))
+            layers.append(
+                self.build_residual_block(self.in_channels, out_channels, stride=1)
+            )
         return nn.Sequential(*layers)
 
     def build_residual_block(self, in_channels, out_channels, stride):
         return nn.Sequential(
-            nn.Conv2d(in_channels, out_channels, kernel_size=3, stride=stride, padding=1, bias=False),
+            nn.Conv2d(
+                in_channels,
+                out_channels,
+                kernel_size=3,
+                stride=stride,
+                padding=1,
+                bias=False,
+            ),
             nn.BatchNorm2d(out_channels),
             nn.ReLU(inplace=True),
-            nn.Conv2d(out_channels, out_channels, kernel_size=3, stride=1, padding=1, bias=False),
+            nn.Conv2d(
+                out_channels,
+                out_channels,
+                kernel_size=3,
+                stride=1,
+                padding=1,
+                bias=False,
+            ),
             nn.BatchNorm2d(out_channels),
         )
 
@@ -237,7 +272,6 @@ def forward(self, x):
         return x
 
 
-
 class ResNet50(nn.Module):
     def __init__(self, num_classes):
         super(ResNet50, self).__init__()
@@ -264,15 +298,31 @@ def make_layer(self, out_channels, num_blocks, stride):
         layers.append(self.build_residual_block(self.in_channels, out_channels, stride))
         self.in_channels = out_channels
         for _ in range(1, num_blocks):
-            layers.append(self.build_residual_block(self.in_channels, out_channels, stride=1))
+            layers.append(
+                self.build_residual_block(self.in_channels, out_channels, stride=1)
+            )
         return nn.Sequential(*layers)
 
     def build_residual_block(self, in_channels, out_channels, stride):
         return nn.Sequential(
-            nn.Conv2d(in_channels, out_channels, kernel_size=3, stride=stride, padding=1, bias=False),
+            nn.Conv2d(
+                in_channels,
+                out_channels,
+                kernel_size=3,
+                stride=stride,
+                padding=1,
+                bias=False,
+            ),
             nn.BatchNorm2d(out_channels),
             nn.ReLU(inplace=True),
-            nn.Conv2d(out_channels, out_channels, kernel_size=3, stride=1, padding=1, bias=False),
+            nn.Conv2d(
+                out_channels,
+                out_channels,
+                kernel_size=3,
+                stride=1,
+                padding=1,
+                bias=False,
+            ),
             nn.BatchNorm2d(out_channels),
         )
 
@@ -317,15 +367,31 @@ def make_layer(self, out_channels, num_blocks, stride):
         layers.append(self.build_residual_block(self.in_channels, out_channels, stride))
         self.in_channels = out_channels
         for _ in range(1, num_blocks):
-            layers.append(self.build_residual_block(self.in_channels, out_channels, stride=1))
+            layers.append(
+                self.build_residual_block(self.in_channels, out_channels, stride=1)
+            )
         return nn.Sequential(*layers)
 
     def build_residual_block(self, in_channels, out_channels, stride):
         return nn.Sequential(
-            nn.Conv2d(in_channels, out_channels, kernel_size=3, stride=stride, padding=1, bias=False),
+            nn.Conv2d(
+                in_channels,
+                out_channels,
+                kernel_size=3,
+                stride=stride,
+                padding=1,
+                bias=False,
+            ),
             nn.BatchNorm2d(out_channels),
             nn.ReLU(inplace=True),
-            nn.Conv2d(out_channels, out_channels, kernel_size=3, stride=1, padding=1, bias=False),
+            nn.Conv2d(
+                out_channels,
+                out_channels,
+                kernel_size=3,
+                stride=1,
+                padding=1,
+                bias=False,
+            ),
             nn.BatchNorm2d(out_channels),
         )
 
@@ -350,7 +416,9 @@ def __init__(self, num_classes, growth_rate=12, num_blocks=3, num_layers=4):
         self.in_channels = 64
 
         # Initial convolution layer
-        self.conv1 = nn.Conv2d(3, self.in_channels, kernel_size=3, stride=1, padding=1, bias=False)
+        self.conv1 = nn.Conv2d(
+            3, self.in_channels, kernel_size=3, stride=1, padding=1, bias=False
+        )
         self.bn1 = nn.BatchNorm2d(self.in_channels)
         self.relu = nn.ReLU(inplace=True)
 
@@ -367,11 +435,20 @@ def make_dense_block(self, growth_rate, num_layers):
         layers = []
         in_channels = self.in_channels
         for _ in range(num_layers):
-            layers.extend([
-                nn.BatchNorm2d(in_channels),
-                nn.ReLU(inplace=True),
-                nn.Conv2d(in_channels, growth_rate, kernel_size=3, stride=1, padding=1, bias=False),
-            ])
+            layers.extend(
+                [
+                    nn.BatchNorm2d(in_channels),
+                    nn.ReLU(inplace=True),
+                    nn.Conv2d(
+                        in_channels,
+                        growth_rate,
+                        kernel_size=3,
+                        stride=1,
+                        padding=1,
+                        bias=False,
+                    ),
+                ]
+            )
             in_channels += growth_rate
         self.in_channels = in_channels
         return nn.Sequential(*layers)
@@ -387,15 +464,28 @@ def forward(self, x):
         x = self.fc(x)
         return x
 
+
 class LightVisionTransformer(nn.Module):
-    def __init__(self, image_size=224, patch_size=16, num_classes=1000, dim=256, depth=6, heads=4, mlp_dim=512, dropout=0.1):
+    def __init__(
+        self,
+        image_size=224,
+        patch_size=16,
+        num_classes=1000,
+        dim=256,
+        depth=6,
+        heads=4,
+        mlp_dim=512,
+        dropout=0.1,
+    ):
         super(LightVisionTransformer, self).__init()
 
         num_patches = (image_size // patch_size) ** 2
         patch_dim = 3 * patch_size * patch_size  # 3 for RGB channels
 
         # Patch embedding layer
-        self.patch_embedding = nn.Conv2d(3, dim, kernel_size=patch_size, stride=patch_size)
+        self.patch_embedding = nn.Conv2d(
+            3, dim, kernel_size=patch_size, stride=patch_size
+        )
         self.positional_embedding = nn.Parameter(torch.randn(1, num_patches + 1, dim))
         self.cls_token = nn.Parameter(torch.randn(1, 1, dim))
         self.transformer = nn.Transformer(
@@ -411,10 +501,11 @@ def forward(self, x):
         B, C, H, W = x.shape
         x = self.patch_embedding(x)
         x = x.permute(0, 2, 3, 1).view(B, -1, x.size(1))  # Flatten and transpose
-        x = torch.cat([self.cls_token.expand(B, -1, -1), x], dim=1)  # Prepend the classification token
+        x = torch.cat(
+            [self.cls_token.expand(B, -1, -1), x], dim=1
+        )  # Prepend the classification token
         x = x + self.positional_embedding
         x = self.transformer(x)
         x = x.mean(dim=1)  # Global average pooling
         x = self.fc(x)
         return x
-
diff --git a/UniTrain/models/segmentation.py b/UniTrain/models/segmentation.py
index aec8015..2570828 100644
--- a/UniTrain/models/segmentation.py
+++ b/UniTrain/models/segmentation.py
@@ -244,10 +244,11 @@ def forward(self, x):
 
         return out
 
+
 class VGGNet(nn.Module):
     def __init__(self, n_class):
         super(VGGNet, self).__init()
-        
+
         # Encoder
         # VGG-like convolutional blocks
         self.encoder = nn.Sequential(
@@ -256,31 +257,27 @@ def __init__(self, n_class):
             nn.Conv2d(64, 64, kernel_size=3, padding=1),
             nn.ReLU(inplace=True),
             nn.MaxPool2d(kernel_size=2, stride=2),
-            
             nn.Conv2d(64, 128, kernel_size=3, padding=1),
             nn.ReLU(inplace=True),
             nn.Conv2d(128, 128, kernel_size=3, padding=1),
             nn.ReLU(inplace=True),
             nn.MaxPool2d(kernel_size=2, stride=2),
-            
             nn.Conv2d(128, 256, kernel_size=3, padding=1),
             nn.ReLU(inplace=True),
             nn.Conv2d(256, 256, kernel_size=3, padding=1),
             nn.ReLU(inplace=True),
             nn.MaxPool2d(kernel_size=2, stride=2),
-            
             nn.Conv2d(256, 512, kernel_size=3, padding=1),
             nn.ReLU(inplace=True),
             nn.Conv2d(512, 512, kernel_size=3, padding=1),
             nn.ReLU(inplace=True),
             nn.MaxPool2d(kernel_size=2, stride=2),
-            
             nn.Conv2d(512, 1024, kernel_size=3, padding=1),
             nn.ReLU(inplace=True),
             nn.Conv2d(1024, 1024, kernel_size=3, padding=1),
-            nn.ReLU(inplace=True)
+            nn.ReLU(inplace=True),
         )
-        
+
         # Decoder
         self.decoder = nn.Sequential(
             nn.ConvTranspose2d(1024, 512, kernel_size=2, stride=2),
@@ -288,37 +285,33 @@ def __init__(self, n_class):
             nn.ReLU(inplace=True),
             nn.Conv2d(512, 512, kernel_size=3, padding=1),
             nn.ReLU(inplace=True),
-            
             nn.ConvTranspose2d(512, 256, kernel_size=2, stride=2),
             nn.Conv2d(512, 256, kernel_size=3, padding=1),
             nn.ReLU(inplace=True),
             nn.Conv2d(256, 256, kernel_size=3, padding=1),
             nn.ReLU(inplace=True),
-            
             nn.ConvTranspose2d(256, 128, kernel_size=2, stride=2),
             nn.Conv2d(256, 128, kernel_size=3, padding=1),
             nn.ReLU(inplace=True),
             nn.Conv2d(128, 128, kernel_size=3, padding=1),
             nn.ReLU(inplace=True),
-            
             nn.ConvTranspose2d(128, 64, kernel_size=2, stride=2),
             nn.Conv2d(128, 64, kernel_size=3, padding=1),
             nn.ReLU(inplace=True),
             nn.Conv2d(64, 64, kernel_size=3, padding=1),
-            nn.ReLU(inplace=True)
+            nn.ReLU(inplace=True),
         )
         # Output layer
         self.output_layer = nn.Conv2d(64, n_class, kernel_size=1)
-    
+
     def forward(self, x):
         # Pass input through the encoder
         enc = self.encoder(x)
-        
+
         # Pass the encoded features through the decoder
         dec = self.decoder(enc)
-        
+
         # Pass the decoder output through the final output layer
         out = self.output_layer(dec)
-        
-        return out
 
+        return out
diff --git a/UniTrain/utils/DCGAN.py b/UniTrain/utils/DCGAN.py
index 162bab0..22c62ed 100644
--- a/UniTrain/utils/DCGAN.py
+++ b/UniTrain/utils/DCGAN.py
@@ -176,7 +176,6 @@ def train_model(
     logger=None,
     iou=False,
 ):
-
     os.makedirs(checkpoint_dir + "/discriminator_checkpoint", exist_ok=True)
     os.makedirs(checkpoint_dir + "/generator_checkpoint", exist_ok=True)
 
@@ -213,20 +212,40 @@ def train_model(
                 device="cpu",
             )
 
-        progress_bar = tqdm(train_data_loader, desc=f'Epoch {epoch + 1}/{epochs}', leave=False, dynamic_ncols=True)
+        progress_bar = tqdm(
+            train_data_loader,
+            desc=f"Epoch {epoch + 1}/{epochs}",
+            leave=False,
+            dynamic_ncols=True,
+        )
 
         for real_images, _ in progress_bar:
             # Train discriminator
             i += 1
-            loss_d, real_score, fake_score = train_discriminator(discriminator_model, generator_model, real_images, opt_d, 128, 128, device='cpu')
+            loss_d, real_score, fake_score = train_discriminator(
+                discriminator_model,
+                generator_model,
+                real_images,
+                opt_d,
+                128,
+                128,
+                device="cpu",
+            )
 
             # Train generator
             loss_g = train_generator(
                 opt_g, discriminator_model, generator_model, batch_size, device="cpu"
             )
 
-            progress_bar.set_postfix({'Loss D': loss_d, 'Loss G': loss_g, 'Real Score': real_score, 'Fake Score': fake_score})
-        
+            progress_bar.set_postfix(
+                {
+                    "Loss D": loss_d,
+                    "Loss G": loss_g,
+                    "Real Score": real_score,
+                    "Fake Score": fake_score,
+                }
+            )
+
         progress_bar.close()
 
         # Record losses & scores
@@ -239,8 +258,6 @@ def train_model(
         save_samples(epoch + epoch, generator_model, fixed_latent, show=False)
 
 
-
-
 def evaluate_model(discriminator_model, dataloader):
     discriminator_model.eval()  # Set the model to evaluation mode
     correct = 0
diff --git a/UniTrain/utils/StyleTransfer.py b/UniTrain/utils/StyleTransfer.py
index add02bf..4af2eab 100644
--- a/UniTrain/utils/StyleTransfer.py
+++ b/UniTrain/utils/StyleTransfer.py
@@ -8,8 +8,10 @@
 
 
 class ContentLoss(nn.Module):
-
-    def __init__(self, target,):
+    def __init__(
+        self,
+        target,
+    ):
         super(ContentLoss, self).__init__()
         # we 'detach' the target content from the tree used
         # to dynamically compute the gradient: this is a stated value,
@@ -20,9 +22,9 @@ def __init__(self, target,):
     def forward(self, input):
         self.loss = F.mse_loss(input, self.target)
         return input
-    
-class StyleLoss(nn.Module):
 
+
+class StyleLoss(nn.Module):
     def __init__(self, target_feature):
         super(StyleLoss, self).__init__()
         self.target = self.gram_matrix(target_feature).detach()
@@ -31,7 +33,7 @@ def forward(self, input):
         G = self.gram_matrix(input)
         self.loss = F.mse_loss(G, self.target)
         return input
-    
+
     def gram_matrix(self, input):
         a, b, c, d = input.size()  # a=batch size(=1)
         # b=number of feature maps
@@ -45,6 +47,7 @@ def gram_matrix(self, input):
         # by dividing by the number of element in each feature maps.
         return G.div(a * b * c * d)
 
+
 # create a module to normalize input image so we can easily put it in a
 # ``nn.Sequential``
 class Normalization(nn.Module):
@@ -56,11 +59,11 @@ def __init__(self, mean, std):
         self.mean = mean.clone().detach().view(-1, 1, 1)
         self.std = std.clone().detach().view(-1, 1, 1)
 
-
     def forward(self, img):
         # normalize ``img``
         return (img - self.mean) / self.std
 
+
 def parse_folder(dataset_path):
     print(dataset_path)
     print(os.getcwd())
@@ -68,8 +71,8 @@ def parse_folder(dataset_path):
     try:
         if os.path.exists(dataset_path):
             # Store paths to train, test, and eval folders if they exist
-            content_path = os.path.join(dataset_path, 'images', 'content')
-            style_path = os.path.join(dataset_path, 'images', 'style')
+            content_path = os.path.join(dataset_path, "images", "content")
+            style_path = os.path.join(dataset_path, "images", "style")
 
             if os.path.exists(content_path) & os.path.exists(style_path):
                 print("Content Data folder path:", content_path)
@@ -86,22 +89,24 @@ def parse_folder(dataset_path):
             else:
                 print("Either content or style directory does not exist")
                 return None
-            
+
         else:
-            print(f"The '{dataset_path}' folder does not exist in the current directory.")
+            print(
+                f"The '{dataset_path}' folder does not exist in the current directory."
+            )
             return None
-        
+
     except Exception as e:
         print("An error occurred:", str(e))
         return None
 
+
 def image_loader(image_name, device):
     # desired size of the output image
     imsize = 512 if torch.cuda.is_available() else 128  # use small size if no GPU
 
-    loader = transforms.Compose([
-        transforms.Resize(imsize),  # scale imported image
-        transforms.ToTensor()]
+    loader = transforms.Compose(
+        [transforms.Resize(imsize), transforms.ToTensor()]  # scale imported image
     )  # transform it into a torch tensor
 
     image = Image.open(image_name)
@@ -111,11 +116,16 @@ def image_loader(image_name, device):
 
     return image.to(device, torch.float)
 
-def get_style_model_and_losses(style_img, content_img, cnn, normalization_mean, normalization_std, 
-                               content_layers = ['conv_4'], # desired depth layers to compute style/content losses
-                               style_layers = ['conv_1', 'conv_2', 'conv_3', 'conv_4', 'conv_5']   
-                               ):
 
+def get_style_model_and_losses(
+    style_img,
+    content_img,
+    cnn,
+    normalization_mean,
+    normalization_std,
+    content_layers=["conv_4"],  # desired depth layers to compute style/content losses
+    style_layers=["conv_1", "conv_2", "conv_3", "conv_4", "conv_5"],
+):
     # normalization module
     normalization = Normalization(normalization_mean, normalization_std)
 
@@ -132,19 +142,19 @@ def get_style_model_and_losses(style_img, content_img, cnn, normalization_mean,
     for layer in cnn.children():
         if isinstance(layer, nn.Conv2d):
             i += 1
-            name = f'conv_{i}'
+            name = f"conv_{i}"
         elif isinstance(layer, nn.ReLU):
-            name = f'relu_{i}'
+            name = f"relu_{i}"
             # The in-place version doesn't play very nicely with the ``ContentLoss``
             # and ``StyleLoss`` we insert below. So we replace with out-of-place
             # ones here.
             layer = nn.ReLU(inplace=False)
         elif isinstance(layer, nn.MaxPool2d):
-            name = f'pool_{i}'
+            name = f"pool_{i}"
         elif isinstance(layer, nn.BatchNorm2d):
-            name = f'bn_{i}'
+            name = f"bn_{i}"
         else:
-            raise RuntimeError(f'Unrecognized layer: {layer.__class__.__name__}')
+            raise RuntimeError(f"Unrecognized layer: {layer.__class__.__name__}")
 
         model.add_module(name, layer)
 
@@ -167,18 +177,31 @@ def get_style_model_and_losses(style_img, content_img, cnn, normalization_mean,
         if isinstance(model[i], ContentLoss) or isinstance(model[i], StyleLoss):
             break
 
-    model = model[:(i + 1)]
+    model = model[: (i + 1)]
 
     return model, style_losses, content_losses
 
-def run_style_transfer(cnn, content_img, style_img, input_img, normalization_mean = torch.tensor([0.485, 0.456, 0.406]), # VGG networks are trained on images with each channel normalized by 
-                        normalization_std = torch.tensor([0.229, 0.224, 0.225]),  # mean=[0.485, 0.456, 0.406] and std=[0.229, 0.224, 0.225]
-                        num_steps=300,
-                       style_weight=1000000, content_weight=1):
+
+def run_style_transfer(
+    cnn,
+    content_img,
+    style_img,
+    input_img,
+    normalization_mean=torch.tensor(
+        [0.485, 0.456, 0.406]
+    ),  # VGG networks are trained on images with each channel normalized by
+    normalization_std=torch.tensor(
+        [0.229, 0.224, 0.225]
+    ),  # mean=[0.485, 0.456, 0.406] and std=[0.229, 0.224, 0.225]
+    num_steps=300,
+    style_weight=1000000,
+    content_weight=1,
+):
     """Run the style transfer."""
-    print('Building the style transfer model..')
-    model, style_losses, content_losses = get_style_model_and_losses(style_img, content_img, cnn,
-        normalization_mean, normalization_std)
+    print("Building the style transfer model..")
+    model, style_losses, content_losses = get_style_model_and_losses(
+        style_img, content_img, cnn, normalization_mean, normalization_std
+    )
 
     # We want to optimize the input and not the model parameters so we
     # update all the requires_grad fields accordingly
@@ -190,7 +213,7 @@ def run_style_transfer(cnn, content_img, style_img, input_img, normalization_mea
 
     optimizer = optim.LBFGS([input_img])
 
-    print('Optimizing..')
+    print("Optimizing..")
     run = [0]
     while run[0] <= num_steps:
 
@@ -218,8 +241,11 @@ def closure():
             run[0] += 1
             if run[0] % 50 == 0:
                 print("run {}:".format(run))
-                print('Style Loss : {:4f} Content Loss: {:4f}'.format(
-                    style_score.item(), content_score.item()))
+                print(
+                    "Style Loss : {:4f} Content Loss: {:4f}".format(
+                        style_score.item(), content_score.item()
+                    )
+                )
                 print()
 
             return style_score + content_score
@@ -232,11 +258,12 @@ def closure():
 
     return input_img
 
-def imshow(tensor, unloader = transforms.ToPILImage(), title=None):
+
+def imshow(tensor, unloader=transforms.ToPILImage(), title=None):
     image = tensor.cpu().clone()  # we clone the tensor to not do changes on it
-    image = image.squeeze(0)      # remove the fake batch dimension
+    image = image.squeeze(0)  # remove the fake batch dimension
     image = unloader(image)
     plt.imshow(image)
     if title is not None:
         plt.title(title)
-    plt.pause(0.001) # pause a bit so that plots are updated
+    plt.pause(0.001)  # pause a bit so that plots are updated
diff --git a/UniTrain/utils/StyleTransferVGG.py b/UniTrain/utils/StyleTransferVGG.py
index e0bcdb3..0b3907c 100644
--- a/UniTrain/utils/StyleTransferVGG.py
+++ b/UniTrain/utils/StyleTransferVGG.py
@@ -1,4 +1,4 @@
-#importing the required libraries
+# importing the required libraries
 import torch
 import torchvision.transforms as transforms
 from PIL import Image
@@ -9,79 +9,95 @@
 import matplotlib.pyplot as plt
 
 
-#Loading the model vgg19 that will serve as the base model
+# Loading the model vgg19 that will serve as the base model
 model = models.vgg19(pretrained=True).features
 # the vgg19 model has three components :
-    #features: Containg all the conv, relu and maxpool
-    #avgpool: Containing the avgpool layer
-    #classifier: Contains the Dense layer(FC part of the model)
+# features: Containg all the conv, relu and maxpool
+# avgpool: Containing the avgpool layer
+# classifier: Contains the Dense layer(FC part of the model)
 
 
-#Assigning the GPU to the variable device
-device=torch.device( "cuda" if (torch.cuda.is_available()) else 'cpu')
+# Assigning the GPU to the variable device
+device = torch.device("cuda" if (torch.cuda.is_available()) else "cpu")
 
-#Defining a class that for the model
+
+# Defining a class that for the model
 class VGG(nn.Module):
     def __init__(self):
-        super(VGG,self).__init__()
-        #Here we will use the following layers and make an array of their indices
+        super(VGG, self).__init__()
+        # Here we will use the following layers and make an array of their indices
         # 0: block1_conv1
         # 5: block2_conv1
         # 10: block3_conv1
         # 19: block4_conv1
         # 28: block5_conv1
-        self.req_features= ['0','5','10','19','28']
-        #Since we need only the 5 layers in the model so we will be dropping all the rest layers from the features of the model
-        self.model=models.vgg19(pretrained=True).features[:29] #model will contain the first 29 layers
-
-
-    #x holds the input tensor(image) that will be fed to each layer
-    def forward(self,x):
-        #initialize an array that wil hold the activations from the chosen layers
-        features=[]
-        #Iterate over all the layers of the mode
-        for layer_num,layer in enumerate(self.model):
-            #activation of the layer will stored in x
-            x=layer(x)
-            #appending the activation of the selected layers and return the feature array
-            if (str(layer_num) in self.req_features):
+        self.req_features = ["0", "5", "10", "19", "28"]
+        # Since we need only the 5 layers in the model so we will be dropping all the rest layers from the features of the model
+        self.model = models.vgg19(pretrained=True).features[
+            :29
+        ]  # model will contain the first 29 layers
+
+    # x holds the input tensor(image) that will be fed to each layer
+    def forward(self, x):
+        # initialize an array that wil hold the activations from the chosen layers
+        features = []
+        # Iterate over all the layers of the mode
+        for layer_num, layer in enumerate(self.model):
+            # activation of the layer will stored in x
+            x = layer(x)
+            # appending the activation of the selected layers and return the feature array
+            if str(layer_num) in self.req_features:
                 features.append(x)
 
         return features
 
 
-
-#defing a function that will load the image and perform the required preprocessing and put it on the GPU
+# defing a function that will load the image and perform the required preprocessing and put it on the GPU
 def image_loader(path):
-    image=Image.open(path)
-    #defining the image transformation steps to be performed before feeding them to the model
-    loader=transforms.Compose([transforms.Resize((512,512)),transforms.ToTensor()])
-    #The preprocessing steps involves resizing the image and then converting it to a tensor
+    image = Image.open(path)
+    # defining the image transformation steps to be performed before feeding them to the model
+    loader = transforms.Compose([transforms.Resize((512, 512)), transforms.ToTensor()])
+    # The preprocessing steps involves resizing the image and then converting it to a tensor
+
+    image = loader(image).unsqueeze(0)
+    return image.to(device, torch.float)
 
-    image=loader(image).unsqueeze(0)
-    return image.to(device,torch.float)
 
-def calc_content_loss(gen_feat,orig_feat):
-    #calculating the content loss of each layer by calculating the MSE between the content and generated features and adding it to content loss
-    content_l=torch.mean((gen_feat-orig_feat)**2)#*0.5
+def calc_content_loss(gen_feat, orig_feat):
+    # calculating the content loss of each layer by calculating the MSE between the content and generated features and adding it to content loss
+    content_l = torch.mean((gen_feat - orig_feat) ** 2)  # *0.5
     return content_l
 
-def calc_style_loss(gen,style):
-    #Calculating the gram matrix for the style and the generated image
-    batch_size,channel,height,width=gen.shape
 
-    G=torch.mm(gen.view(channel,height*width),gen.view(channel,height*width).t())
-    A=torch.mm(style.view(channel,height*width),style.view(channel,height*width).t())
+def calc_style_loss(gen, style):
+    # Calculating the gram matrix for the style and the generated image
+    batch_size, channel, height, width = gen.shape
 
-    #Calcultating the style loss of each layer by calculating the MSE between the gram matrix of the style image and the generated image and adding it to style loss
-    style_l=torch.mean((G-A)**2)#/(4*channel*(height*width)**2)
+    G = torch.mm(
+        gen.view(channel, height * width), gen.view(channel, height * width).t()
+    )
+    A = torch.mm(
+        style.view(channel, height * width), style.view(channel, height * width).t()
+    )
+
+    # Calcultating the style loss of each layer by calculating the MSE between the gram matrix of the style image and the generated image and adding it to style loss
+    style_l = torch.mean((G - A) ** 2)  # /(4*channel*(height*width)**2)
     return style_l
 
-def train_style_transfer_model(original_image_path, style_image_path, final_output_path, epoch = 1000, optimizer = optim.Adam, lr = 0.004, alpha = 8, beta = 70):
-  
-  '''Train a PyTorch model for a style transfer task.
+
+def train_style_transfer_model(
+    original_image_path,
+    style_image_path,
+    final_output_path,
+    epoch=1000,
+    optimizer=optim.Adam,
+    lr=0.004,
+    alpha=8,
+    beta=70,
+):
+    """Train a PyTorch model for a style transfer task.
     Args:
-    
+
     original_image_path (path): Path on your device/drive where original image is stored.
     style_image_path (path): Path on your device/drive where style image is stored.
     final_output_path (path): Path on your device/drive where output image is to be saved.
@@ -93,48 +109,47 @@ def train_style_transfer_model(original_image_path, style_image_path, final_outp
 
     Returns:
     None
-    '''
-  
-
-  #Loading the original and the style image
-  original_image = image_loader(original_image_path)
-  style_image = image_loader(style_image_path)
-
-  #Creating the generated image from the original image
-  generated_image = original_image.clone().requires_grad_(True)
-
-  #Load the model to the GPU
-  model=VGG().to(device).eval()
-
-  #using adam optimizer and it will update the generated image not the model parameter
-  optimizer = optimizer([generated_image],lr=lr)
-
-  def calculate_loss(gen_features, orig_feautes, style_featues):
-    style_loss=content_loss=0
-    for gen,cont,style in zip(gen_features,orig_feautes,style_featues):
-        #extracting the dimensions from the generated image
-        content_loss+=calc_content_loss(gen,cont)
-        style_loss+=calc_style_loss(gen,style)
-
-    #calculating the total loss of e th epoch
-    total_loss=alpha*content_loss + beta*style_loss
-    return total_loss
-
-  for e in range (epoch):
-      #extracting the features of generated, content and the original required for calculating the loss
-      gen_features=model(generated_image)
-      orig_feautes=model(original_image)
-      style_featues=model(style_image)
-
-      #iterating over the activation of each layer and calculate the loss and add it to the content and the style loss
-      total_loss=calculate_loss(gen_features, orig_feautes, style_featues)
-      #optimize the pixel values of the generated image and backpropagate the loss
-      optimizer.zero_grad()
-      total_loss.backward()
-      optimizer.step()
-
-      #print the image and save it after each 100 epoch
-      if(not (e%100)):
-          print(total_loss)
-
-          save_image(generated_image,final_output_path)
\ No newline at end of file
+    """
+
+    # Loading the original and the style image
+    original_image = image_loader(original_image_path)
+    style_image = image_loader(style_image_path)
+
+    # Creating the generated image from the original image
+    generated_image = original_image.clone().requires_grad_(True)
+
+    # Load the model to the GPU
+    model = VGG().to(device).eval()
+
+    # using adam optimizer and it will update the generated image not the model parameter
+    optimizer = optimizer([generated_image], lr=lr)
+
+    def calculate_loss(gen_features, orig_feautes, style_featues):
+        style_loss = content_loss = 0
+        for gen, cont, style in zip(gen_features, orig_feautes, style_featues):
+            # extracting the dimensions from the generated image
+            content_loss += calc_content_loss(gen, cont)
+            style_loss += calc_style_loss(gen, style)
+
+        # calculating the total loss of e th epoch
+        total_loss = alpha * content_loss + beta * style_loss
+        return total_loss
+
+    for e in range(epoch):
+        # extracting the features of generated, content and the original required for calculating the loss
+        gen_features = model(generated_image)
+        orig_feautes = model(original_image)
+        style_featues = model(style_image)
+
+        # iterating over the activation of each layer and calculate the loss and add it to the content and the style loss
+        total_loss = calculate_loss(gen_features, orig_feautes, style_featues)
+        # optimize the pixel values of the generated image and backpropagate the loss
+        optimizer.zero_grad()
+        total_loss.backward()
+        optimizer.step()
+
+        # print the image and save it after each 100 epoch
+        if not (e % 100):
+            print(total_loss)
+
+            save_image(generated_image, final_output_path)
diff --git a/UniTrain/utils/__init__.py b/UniTrain/utils/__init__.py
index ba441da..84c7d89 100644
--- a/UniTrain/utils/__init__.py
+++ b/UniTrain/utils/__init__.py
@@ -1,4 +1,9 @@
 from .classification import get_data_loader, parse_folder, train_model
 from .DCGAN import get_data_loader, parse_folder, train_model
-from .segmentation import (generate_model_summary, get_data_loader, iou_score,
-                           parse_folder, train_unet)
+from .segmentation import (
+    generate_model_summary,
+    get_data_loader,
+    iou_score,
+    parse_folder,
+    train_unet,
+)
diff --git a/UniTrain/utils/classification.py b/UniTrain/utils/classification.py
index 98924cf..5260c49 100644
--- a/UniTrain/utils/classification.py
+++ b/UniTrain/utils/classification.py
@@ -10,7 +10,8 @@
 import tqdm
 from PIL import Image
 
-def get_data_loader(data_dir, batch_size, shuffle=True, transform = None, split='train'):
+
+def get_data_loader(data_dir, batch_size, shuffle=True, transform=None, split="train"):
     """
     Create and return a data loader for a custom dataset.
 
@@ -24,33 +25,31 @@ def get_data_loader(data_dir, batch_size, shuffle=True, transform = None, split=
     """
     # Define data transformations (adjust as needed)
 
-    if split == 'train':
-        data_dir = os.path.join(data_dir, 'train')
-    elif split == 'test':
-        data_dir = os.path.join(data_dir, 'test')
-    elif split == 'eval':
-        data_dir = os.path.join(data_dir, 'eval')
+    if split == "train":
+        data_dir = os.path.join(data_dir, "train")
+    elif split == "test":
+        data_dir = os.path.join(data_dir, "test")
+    elif split == "eval":
+        data_dir = os.path.join(data_dir, "eval")
     else:
         raise ValueError(f"Invalid split choice: {split}")
 
-
-
     if transform is None:
-        transform = transforms.Compose([
-            transforms.Resize((224, 224)),  # Resize images to a fixed size
-            transforms.ToTensor(),  # Convert images to PyTorch tensors
-            transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))  # Normalize with ImageNet stats
-        ])
+        transform = transforms.Compose(
+            [
+                transforms.Resize((224, 224)),  # Resize images to a fixed size
+                transforms.ToTensor(),  # Convert images to PyTorch tensors
+                transforms.Normalize(
+                    (0.485, 0.456, 0.406), (0.229, 0.224, 0.225)
+                ),  # Normalize with ImageNet stats
+            ]
+        )
 
     # Create a custom dataset
     dataset = ClassificationDataset(data_dir, transform=transform)
 
     # Create a data loader
-    data_loader = DataLoader(
-        dataset,
-        batch_size=batch_size,
-        shuffle=shuffle
-    )
+    data_loader = DataLoader(dataset, batch_size=batch_size, shuffle=shuffle)
 
     return data_loader
 
@@ -59,11 +58,15 @@ def parse_folder(dataset_path):
     try:
         if os.path.exists(dataset_path):
             # Store paths to train, test, and eval folders if they exist
-            train_path = os.path.join(dataset_path, 'train')
-            test_path = os.path.join(dataset_path, 'test')
-            eval_path = os.path.join(dataset_path, 'eval')
-
-            if os.path.exists(train_path) and os.path.exists(test_path) and os.path.exists(eval_path):
+            train_path = os.path.join(dataset_path, "train")
+            test_path = os.path.join(dataset_path, "test")
+            eval_path = os.path.join(dataset_path, "eval")
+
+            if (
+                os.path.exists(train_path)
+                and os.path.exists(test_path)
+                and os.path.exists(eval_path)
+            ):
                 print("Train folder path:", train_path)
                 print("Test folder path:", test_path)
                 print("Eval folder path:", eval_path)
@@ -81,16 +84,29 @@ def parse_folder(dataset_path):
                 print("One or more of the train, test, or eval folders does not exist.")
                 return None
         else:
-            print(f"The '{dataset_path}' folder does not exist in the current directory.")
+            print(
+                f"The '{dataset_path}' folder does not exist in the current directory."
+            )
             return None
     except Exception as e:
         print("An error occurred:", str(e))
         return None
 
-def train_model(model, train_data_loader, test_data_loader, num_epochs, learning_rate=0.001, criterion_fn = nn.CrossEntropyLoss, optimizer_fn = optim.Adam, checkpoint_dir='checkpoints', wnb_dir='wnb', logger=None, device=torch.device('cpu')):
 
-
-    '''Train a PyTorch model for a classification task.
+def train_model(
+    model,
+    train_data_loader,
+    test_data_loader,
+    num_epochs,
+    learning_rate=0.001,
+    criterion_fn=nn.CrossEntropyLoss,
+    optimizer_fn=optim.Adam,
+    checkpoint_dir="checkpoints",
+    wnb_dir="wnb",
+    logger=None,
+    device=torch.device("cpu"),
+):
+    """Train a PyTorch model for a classification task.
     Args:
     model (nn.Module): Torch model to train.
     train_data_loader (DataLoader): Training data loader.
@@ -108,20 +124,21 @@ def train_model(model, train_data_loader, test_data_loader, num_epochs, learning
 
     Returns:
     None
-    '''
+    """
 
     if logger:
-        logging.basicConfig(level=logging.INFO, format='%(asctime)s - Epoch %(epoch)d - Train Acc: %(train_acc).4f - Val Acc: %(val_acc).4f - %(message)s',
-                        datefmt='%Y-%m-%d %H:%M:%S', filename=logger, filemode='w')
+        logging.basicConfig(
+            level=logging.INFO,
+            format="%(asctime)s - Epoch %(epoch)d - Train Acc: %(train_acc).4f - Val Acc: %(val_acc).4f - %(message)s",
+            datefmt="%Y-%m-%d %H:%M:%S",
+            filename=logger,
+            filemode="w",
+        )
         logger = logging.getLogger(__name__)
 
-
-
     # Setting the optimizer and criterion
     optimizer = optimizer_fn(model.parameters(), lr=learning_rate)
     criterion = criterion_fn()
-    
-
 
     # Initialize optimizer, loss and accuracy
     optimizer = optimizer(model.parameters(), lr=learning_rate)
@@ -150,34 +167,41 @@ def train_model(model, train_data_loader, test_data_loader, num_epochs, learning
 
             running_loss += loss.item()
             loop.set_description(f"Epoch [{epoch+1}/{num_epochs}]")
-            loop.set_postfix(loss= running_loss / (batch_idx + 1))
+            loop.set_postfix(loss=running_loss / (batch_idx + 1))
             if batch_idx % 100 == 99:  # Print and log every 100 batches
                 avg_loss = running_loss / 100
-                
+
                 # Save the weights and biases and log the path.
-                wnb_path = os.path.join(wnb_dir, f'model_epoch_{epoch + 1}_batch{batch_idx + 1}.pth')
+                wnb_path = os.path.join(
+                    wnb_dir, f"model_epoch_{epoch + 1}_batch{batch_idx + 1}.pth"
+                )
                 torch.save(model.state_dict(), wnb_path)
                 if logger:
-                    logger.info(f'Epoch {epoch + 1}, Batch {batch_idx + 1}, Loss: {avg_loss:.4f}, wnbPath: {wnb_path}')
-
+                    logger.info(
+                        f"Epoch {epoch + 1}, Batch {batch_idx + 1}, Loss: {avg_loss:.4f}, wnbPath: {wnb_path}"
+                    )
 
         accuracy = evaluate_model(model, test_data_loader)
-        
+
         # Save the weights and biases and log the path for current epoch.
-        wnb_path = os.path.join(wnb_dir, f'model_epoch_{epoch + 1}.pth')
+        wnb_path = os.path.join(wnb_dir, f"model_epoch_{epoch + 1}.pth")
         torch.save(model.state_dict(), wnb_path)
         if logger:
-            logger.info(f'Epoch {epoch + 1}, Validation Accuracy: {accuracy:.2f}%, wnbPath: {wnb_path}')
+            logger.info(
+                f"Epoch {epoch + 1}, Validation Accuracy: {accuracy:.2f}%, wnbPath: {wnb_path}"
+            )
 
         # Save model checkpoint if accuracy improves
         if accuracy > best_accuracy:
             best_accuracy = accuracy
-            checkpoint_path = os.path.join(checkpoint_dir, f'model_epoch_{epoch + 1}.pth')
+            checkpoint_path = os.path.join(
+                checkpoint_dir, f"model_epoch_{epoch + 1}.pth"
+            )
             torch.save(model.state_dict(), checkpoint_path)
             if logger:
-                logger.info(f'Saved checkpoint to {checkpoint_path}')
+                logger.info(f"Saved checkpoint to {checkpoint_path}")
 
-    print('Finished Training')
+    print("Finished Training")
 
 
 def evaluate_model(model, dataloader):
@@ -197,12 +221,14 @@ def evaluate_model(model, dataloader):
     return accuracy
 
 
-def infer_class(model: nn.Module, image_path: str, device: torch.device, dataloader: DataLoader) -> str:
+def infer_class(
+    model: nn.Module, image_path: str, device: torch.device, dataloader: DataLoader
+) -> str:
     """Perform inference on a single image.
 
     Args:
         model (nn.Module): Model to perform inference with.
-        image_path (str): Path to image to perform inference on.    
+        image_path (str): Path to image to perform inference on.
         device (torch.device): Device to run inference on (GPU or CPU).
         dataloader (DataLoader): Data loader for the dataset.
 
@@ -215,20 +241,21 @@ def infer_class(model: nn.Module, image_path: str, device: torch.device, dataloa
 
     # Define transformations for the image
     if transform is None:
-        transform = transforms.Compose([
-            transforms.Resize((224, 224)),
-            transforms.ToTensor(),
-            transforms.Normalize((0.485, 0.456, 0.406),
-                                 (0.229, 0.224, 0.225))
-        ])
+        transform = transforms.Compose(
+            [
+                transforms.Resize((224, 224)),
+                transforms.ToTensor(),
+                transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225)),
+            ]
+        )
 
-    image = Image.open(image_path).convert('RGB')
+    image = Image.open(image_path).convert("RGB")
 
     image_tensor = transform(image)
 
     # Add an extra batch dimension since pytorch treats all images as batches
     image_tensor = image_tensor.unsqueeze_(0)
-    
+
     with torch.no_grad():
         output = model(image_tensor.to(device))
 
@@ -238,4 +265,3 @@ def infer_class(model: nn.Module, image_path: str, device: torch.device, dataloa
     classes = dataloader.dataset.classes
 
     return classes[predicted]
-
diff --git a/UniTrain/utils/segmentation.py b/UniTrain/utils/segmentation.py
index 2bbfeb5..fc4da1a 100644
--- a/UniTrain/utils/segmentation.py
+++ b/UniTrain/utils/segmentation.py
@@ -102,7 +102,6 @@ def parse_folder(dataset_path):
         return False
 
 
-
 def train_unet(
     model,
     train_data_loader,
@@ -117,23 +116,23 @@ def train_unet(
     device=torch.device("cpu"),
 ) -> None:
     """
-    Args:
-
-def train_unet(model, train_data_loader, test_data_loader, num_epochs, learning_rate, checkpoint_dir, optimizer = optim.Adam, loss_criterion = nn.CrossEntropyLoss, logger=None, iou=False, device=torch.device('cpu')) -> None:
-    '''
-    Args: 
-    model (nn.Module): PyTorch model to train.
-    train_data_loader (DataLoader): Data loader of the training dataset.
-    test_data_loader (DataLoader): Data loader of the test dataset.
-    num_epochs (int): Number of epochs to train the model.
-    learning_rate (float): Learning rate for the optimizer.
-    checkpoint_dir (str): Directory to save model checkpoints.
-    logger (Logger): Logger to log training information.
-    iou (bool): Whether to calculate the IOU score.
-    device (torch.device): Device to run training on (GPU or CPU).
-
-    Returns:
-    None
+        Args:
+
+    def train_unet(model, train_data_loader, test_data_loader, num_epochs, learning_rate, checkpoint_dir, optimizer = optim.Adam, loss_criterion = nn.CrossEntropyLoss, logger=None, iou=False, device=torch.device('cpu')) -> None:
+        '''
+        Args:
+        model (nn.Module): PyTorch model to train.
+        train_data_loader (DataLoader): Data loader of the training dataset.
+        test_data_loader (DataLoader): Data loader of the test dataset.
+        num_epochs (int): Number of epochs to train the model.
+        learning_rate (float): Learning rate for the optimizer.
+        checkpoint_dir (str): Directory to save model checkpoints.
+        logger (Logger): Logger to log training information.
+        iou (bool): Whether to calculate the IOU score.
+        device (torch.device): Device to run training on (GPU or CPU).
+
+        Returns:
+        None
     """
 
     if logger:
@@ -242,26 +241,28 @@ def iou_score(output, target):
     iou = (intersection + smooth) / (union + smooth)
     return iou.mean().item()
 
+
 def do_inference(image: PIL.Image, device: torch.device(), model) -> PIL.Image:
     """
     Function is used for inference for segmentation of an Image.
-    Function takes PIL.Image object as input and return a segmented PIL.Image object.    
-    
+    Function takes PIL.Image object as input and return a segmented PIL.Image object.
+
     Args:
         image(PIL.Image) : Image to do inference
         device(torch.device) : Device to run inference
         model: Model to inference
     """
 
-    model.eval() # Evaulation mode..
+    model.eval()  # Evaulation mode..
 
     # Convert Image.PIL into tensor form.
-    transform = transforms.Compose([ 
-                transforms.Resize((224, 224)),
-                transforms.ToTensor(),
-                transforms.Normalize((0.485, 0.456, 0.406),
-                                        (0.229, 0.224, 0.225))
-                                    ]) 
+    transform = transforms.Compose(
+        [
+            transforms.Resize((224, 224)),
+            transforms.ToTensor(),
+            transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225)),
+        ]
+    )
     image = transform(image)
 
     # Move the image to device and adding extra dimension for batch.
@@ -273,5 +274,7 @@ def do_inference(image: PIL.Image, device: torch.device(), model) -> PIL.Image:
 
     # Post-process the segmentation mask.
     output = output.squeeze(0).cpu().numpy()
-    output = nn.functional.softmax(torch.from_numpy(output), dim=0).argmax(0).cpu().numpy()
+    output = (
+        nn.functional.softmax(torch.from_numpy(output), dim=0).argmax(0).cpu().numpy()
+    )
     return transforms.ToPILImage(output)

From 1772d570af7204e9dc069db6fa8b73eaa42eac87 Mon Sep 17 00:00:00 2001
From: Geet Sethi <geetsethi05@gmail.com>
Date: Mon, 30 Oct 2023 14:27:17 +0530
Subject: [PATCH 2/2] add v-net model #11

---
 UniTrain/models/segmentation.py | 171 ++++++++++++++++++++++++++++++++
 1 file changed, 171 insertions(+)

diff --git a/UniTrain/models/segmentation.py b/UniTrain/models/segmentation.py
index 2570828..778c90d 100644
--- a/UniTrain/models/segmentation.py
+++ b/UniTrain/models/segmentation.py
@@ -315,3 +315,174 @@ def forward(self, x):
         out = self.output_layer(dec)
 
         return out
+
+
+def passthrough(x, **kwargs):
+    return x
+
+
+def ELUCons(elu, nchan):
+    if elu:
+        return nn.ELU(inplace=True)
+    else:
+        return nn.PReLU(nchan)
+
+
+# normalization between sub-volumes is necessary
+# for good performance
+class ContBatchNorm3d(nn.modules.batchnorm._BatchNorm):
+    def _check_input_dim(self, input):
+        if input.dim() != 5:
+            raise ValueError("expected 5D input (got {}D input)".format(input.dim()))
+        super(ContBatchNorm3d, self)._check_input_dim(input)
+
+    def forward(self, input):
+        self._check_input_dim(input)
+        return F.batch_norm(
+            input,
+            self.running_mean,
+            self.running_var,
+            self.weight,
+            self.bias,
+            True,
+            self.momentum,
+            self.eps,
+        )
+
+
+class LUConv(nn.Module):
+    def __init__(self, nchan, elu):
+        super(LUConv, self).__init__()
+        self.relu1 = ELUCons(elu, nchan)
+        self.conv1 = nn.Conv3d(nchan, nchan, kernel_size=5, padding=2)
+        self.bn1 = ContBatchNorm3d(nchan)
+
+    def forward(self, x):
+        out = self.relu1(self.bn1(self.conv1(x)))
+        return out
+
+
+def _make_nConv(nchan, depth, elu):
+    return nn.Sequential(*[LUConv(nchan, elu) for _ in range(depth)])
+
+
+class InputTransition(nn.Module):
+    def __init__(self, outChans, elu):
+        super(InputTransition, self).__init__()
+        self.conv1 = nn.Conv3d(1, 16, kernel_size=5, padding=2)
+        self.bn1 = ContBatchNorm3d(16)
+        self.relu1 = nn.PReLU(16)
+
+    def forward(self, x):
+        out = self.bn1(self.conv1(x))
+        # split input in to 16 channels
+        x16 = torch.cat(tuple(x for _ in range(16)), 0)
+        out = self.relu1(torch.add(out, x16))
+        return out
+
+
+class DownTransition(nn.Module):
+    def __init__(self, inChans, nConvs, elu, dropout=False):
+        super(DownTransition, self).__init__()
+        outChans = 2 * inChans
+        self.down_conv = nn.Conv3d(inChans, outChans, kernel_size=2, stride=2)
+        self.bn1 = ContBatchNorm3d(outChans)
+        self.do1 = passthrough
+        self.relu1 = ELUCons(elu, outChans)
+        self.relu2 = ELUCons(elu, outChans)
+        if dropout:
+            self.do1 = nn.Dropout3d()
+        self.ops = _make_nConv(outChans, nConvs, elu)
+
+    def forward(self, x):
+        down = self.relu1(self.bn1(self.down_conv(x)))
+        out = self.do1(down)
+        out = self.ops(out)
+        out = self.relu2(torch.add(out, down))
+        return out
+
+
+class UpTransition(nn.Module):
+    def __init__(self, inChans, outChans, nConvs, elu, dropout=False):
+        super(UpTransition, self).__init__()
+        self.up_conv = nn.ConvTranspose3d(
+            inChans, outChans // 2, kernel_size=2, stride=2
+        )
+        self.bn1 = ContBatchNorm3d(outChans // 2)
+        self.do1 = passthrough
+        self.do2 = nn.Dropout3d()
+        self.relu1 = ELUCons(elu, outChans // 2)
+        self.relu2 = ELUCons(elu, outChans)
+        if dropout:
+            self.do1 = nn.Dropout3d()
+        self.ops = _make_nConv(outChans, nConvs, elu)
+
+    def forward(self, x, skipx):
+        out = self.do1(x)
+        skipxdo = self.do2(skipx)
+        out = self.relu1(self.bn1(self.up_conv(out)))
+        xcat = torch.cat((out, skipxdo), 1)
+        out = self.ops(xcat)
+        out = self.relu2(torch.add(out, xcat))
+        return out
+
+
+class OutputTransition(nn.Module):
+    def __init__(self, inChans, elu, nll):
+        super(OutputTransition, self).__init__()
+        self.conv1 = nn.Conv3d(inChans, 2, kernel_size=5, padding=2)
+        self.bn1 = ContBatchNorm3d(2)
+        self.conv2 = nn.Conv3d(2, 2, kernel_size=1)
+        self.relu1 = ELUCons(elu, 2)
+        if nll:
+            self.softmax = F.log_softmax
+        else:
+            self.softmax = F.softmax
+
+    def forward(self, x):
+        # convolve 32 down to 2 channels
+        out = self.relu1(self.bn1(self.conv1(x)))
+        out = self.conv2(out)
+
+        # make channels the last axis
+        out = out.permute(0, 2, 3, 4, 1).contiguous()
+        # flatten
+        out = out.view(out.numel() // 2, 2)
+        out = self.softmax(out)
+        # treat channel 0 as the predicted output
+        return out
+
+
+class VNet(nn.Module):
+    def __init__(self, elu=True, nll=False):
+        super(VNet, self).__init__()
+        # Input
+        self.in_tr = InputTransition(16, elu)
+
+        # Encoding
+        self.down_tr32 = DownTransition(16, 1, elu)
+        self.down_tr64 = DownTransition(32, 2, elu)
+        self.down_tr128 = DownTransition(64, 3, elu, dropout=True)
+        self.down_tr256 = DownTransition(128, 2, elu, dropout=True)
+
+        # Decoding
+        self.up_tr256 = UpTransition(256, 256, 2, elu, dropout=True)
+        self.up_tr128 = UpTransition(256, 128, 2, elu, dropout=True)
+        self.up_tr64 = UpTransition(128, 64, 1, elu)
+        self.up_tr32 = UpTransition(64, 32, 1, elu)
+
+        # Output
+        self.out_tr = OutputTransition(32, elu, nll)
+
+    def forward(self, x):
+        in16 = self.in_tr(x)
+        in32 = self.down_tr32(in16)
+        in64 = self.down_tr64(in32)
+        in128 = self.down_tr128(in64)
+        in256 = self.down_tr256(in128)
+        out256 = self.up_tr256(in256, in128)
+        out128 = self.up_tr128(out256, in64)
+        out64 = self.up_tr64(out128, in32)
+        out32 = self.up_tr32(out64, in16)
+        out16 = self.out_tr(out32)
+        return out16