fix pixtral image prompt order for doc VQA (#99)

* fix pixtral image prompt order for doc VQA * fix code style
Blaizzy · Oct 22, 2024 · f0812b5 · f0812b5
1 parent 4da085e
commit f0812b5
Showing 1 changed file with 6 additions and 1 deletion.
diff --git a/mlx_vlm/prompt_utils.py b/mlx_vlm/prompt_utils.py
@@ -24,7 +24,12 @@ def add_image_tokens(message, token_format):
             return message
         if role == "user" and not skip_image_token:
             if isinstance(message["content"], list):
-                message["content"].extend([{"type": "image"}] * num_images)
+                if model_name == "pixtral":
+                    message["content"] = [{"type": "image"}] * num_images + message[
+                        "content"
+                    ]
+                else:
+                    message["content"].extend([{"type": "image"}] * num_images)
             else:
                 if model_name == "phi3_v":
                     message["content"] = f"{token_format}{message['content']}"