Update app.py

app.py

@@ -18,32 +18,37 @@ class PhiModel(nn.Module):
         # NN branch
         self.fc1 = nn.Linear(input_size, 64)
         self.fc2 = nn.Linear(64, 32)
+        self.fc_out = nn.Linear(32, 64)  # To make the NN output size match other branches
 
         # RNN branch
         self.lstm = nn.LSTM(input_size, 64, batch_first=True)
+        self.rnn_fc = nn.Linear(64, 64)  # Adjust RNN output size
 
         # CNN branch
         self.conv1 = nn.Conv1d(1, 32, kernel_size=3, padding=1)
-        self.fc3 = nn.Linear(32 * input_size, 1)
+        self.fc3 = nn.Linear(32 * input_size, 64)  # Adjust CNN output size
 
     def forward(self, nn_input, rnn_input, cnn_input):
         # NN branch
         nn_out = F.relu(self.fc1(nn_input))
         nn_out = F.relu(self.fc2(nn_out))
+        nn_out = self.fc_out(nn_out)  # Ensure the output size matches others
 
         # RNN branch
         rnn_out, _ = self.lstm(rnn_input)
-        rnn_out = rnn_out[:, -1, :]  # take last output of LSTM
+        rnn_out = rnn_out[:, -1, :]  # Take last output of LSTM
+        rnn_out = self.rnn_fc(rnn_out)  # Adjust RNN output size to match NN's output size
 
         # CNN branch
         cnn_input = cnn_input.unsqueeze(1)  # Add channel dimension
         cnn_out = F.relu(self.conv1(cnn_input))
         cnn_out = cnn_out.view(cnn_out.size(0), -1)  # Flatten the output
+        cnn_out = self.fc3(cnn_out)  # Adjust CNN output size to match NN and RNN
 
         # Combine all branches
         combined = nn_out + rnn_out + cnn_out
-        output = torch.sigmoid(self.fc3(combined))
-        
+        output = torch.sigmoid(combined)
+
         return output
 
 # Initialize the model