diff --git a/timm/models/vision_transformer.py b/timm/models/vision_transformer.py
index 1ab2e736..e560ec9a 100644
--- a/timm/models/vision_transformer.py
+++ b/timm/models/vision_transformer.py
@@ -635,13 +635,14 @@ class VisionTransformer(nn.Module):
     ) -> List[torch.Tensor]:
         outputs, num_blocks = [], len(self.blocks)
         take_indices = set(range(num_blocks - n, num_blocks) if isinstance(n, int) else n)
+        last_index_to_take = max(take_indices)
 
         # forward pass
         x = self.patch_embed(x)
         x = self._pos_embed(x)
         x = self.patch_drop(x)
         x = self.norm_pre(x)
-        for i, blk in enumerate(self.blocks):
+        for i, blk in enumerate(self.blocks[: last_index_to_take + 1]):
             x = blk(x)
             if i in take_indices:
                 outputs.append(x)