Update app.py

app.py

@@ -85,21 +85,14 @@ def stream_text_chat(message, history, system_prompt, temperature=0.8, max_new_t
 def process_vision_query(image, text_input):
     prompt = f"<|user|>\n<|image_1|>\n{text_input}<|end|>\n<|assistant|>\n"
     
-    # Convert the image to bytes if it's not already
-    if isinstance(image, Image.Image):
-        # If it's a PIL Image, convert to bytes
-        import io
-        img_byte_arr = io.BytesIO()
-        image.save(img_byte_arr, format='PNG')
-        image = img_byte_arr.getvalue()
-    elif isinstance(image, np.ndarray):
-        # If it's a numpy array, convert to PIL Image first, then to bytes
+    # Ensure the image is in the correct format
+    if isinstance(image, np.ndarray):
+        # Convert numpy array to PIL Image
         image = Image.fromarray(image).convert("RGB")
-        img_byte_arr = io.BytesIO()
-        image.save(img_byte_arr, format='PNG')
-        image = img_byte_arr.getvalue()
+    elif not isinstance(image, Image.Image):
+        raise ValueError("Invalid image type. Expected PIL.Image.Image or numpy.ndarray")
     
-    # Now process the image bytes
+    # Now process the image
     inputs = vision_processor(prompt, images=image, return_tensors="pt").to(device)
     
     with torch.no_grad():