Update cumo/model/multimodal_encoder/clip_encoder.py

cumo/model/multimodal_encoder/clip_encoder.py

@@ -48,7 +48,9 @@ class CLIPVisionTower(nn.Module):
         self.is_loaded = True
 
     def feature_select(self, image_features):
-        #image_features = image_forward_outs.hidden_states[self.select_layer]
+        # Take first element if output is a tuple
+        if isinstance(image_features, tuple):
+            image_features = image_features[0]
         if self.select_feature == 'patch':
             image_features = image_features[:, 1:]
         elif self.select_feature == 'cls_patch':
@@ -57,6 +59,7 @@ class CLIPVisionTower(nn.Module):
             raise ValueError(f'Unexpected select feature: {self.select_feature}')
         return image_features
 
+
     def split_chessboard(self, x, num_split):
         """
             x: b * c * h * w