diff --git a/vllm/model_executor/models/h2ovl.py b/vllm/model_executor/models/h2ovl.py
index 31903f18da9c1..43242fe370ba2 100644
--- a/vllm/model_executor/models/h2ovl.py
+++ b/vllm/model_executor/models/h2ovl.py
@@ -272,10 +272,9 @@ def input_processor(
         elif is_list_of(image_data, torch.Tensor):
 
             image_feature_sizes = []
-            for image in image_data:
-                _, image_feature_size, _ = image.shape
+            for image_embed in image_data:
+                _, image_feature_size, _ = image_embed.shape
                 image_feature_sizes.append(image_feature_size)
-            _, image_feature_size, _ = image_data.shape
             pixel_values = None
 
         else: