src/aiunn/upsampler/aiunn.py

@@ -32,7 +32,18 @@ class aiuNN(PreTrainedModel):
     def forward(self, x):
         if self.base_model is None:
             raise ValueError("Base model is not loaded. Call 'load_base_model' before forwarding.")
-        x = self.base_model(x)  # Get base features
+            
+        # Get base features - we need to extract the last hidden state if it's returned as part of a tuple/dict
+        base_output = self.base_model(x)
+        if isinstance(base_output, tuple):
+            x = base_output[0]
+        elif isinstance(base_output, dict):
+            x = base_output.get('last_hidden_state', base_output.get('hidden_states'))
+            if x is None:
+                raise ValueError("Expected 'last_hidden_state' or 'hidden_states' in model output")
+        else:
+            x = base_output
+            
         x = self.pixel_shuffle_conv(x)  # Expand channels for shuffling
         x = self.pixel_shuffle(x)  # Rearrange channels into spatial dimensions
         return x