Update cross_visual.py

cross_visual.py

@@ -794,4 +794,4 @@ class CrossVisionModel(nn.Module):
 
     def forward(self, images):
         enc = self.vit(images)
-        return enc + self.pos_embed.unsqueeze(0)
+        return enc + self.pos_embed.to(enc.device).unsqueeze(0)