diff --git a/src/axolotl/utils/collators/mm_chat.py b/src/axolotl/utils/collators/mm_chat.py
index 772fe818a..7ec9c77f9 100644
--- a/src/axolotl/utils/collators/mm_chat.py
+++ b/src/axolotl/utils/collators/mm_chat.py
@@ -213,9 +213,12 @@ class MultiModalChatDataCollator(DataCollatorMixin):
                 for example in examples
             ]
 
-        if chat_template_type == "llava":
-            max_images = 1
-        images = __class__.process_images(examples, max_images=max_images)
+        if chat_template_type != "llava":
+            images = __class__.process_images(examples, max_images=max_images)
+        else:
+            # LLava1.5 does not support multiple images
+            images = [example["images"] for example in examples]
+            images = [image[0] for image in images]
 
         # Tokenize the texts and process the images
         batch = processor(text=texts, images=images, return_tensors="pt", padding=True)