Update code files

app.py

@@ -3,14 +3,13 @@ from transformers import Qwen2VLForConditionalGeneration, AutoTokenizer, AutoPro
 import torch
 from qwen_vl_utils import process_vision_info
 from PIL import Image
-import os
 import re
 import gradio as gr
 
 rag = RAGMultiModalModel.from_pretrained("vidore/colpali")
 vlm = Qwen2VLForConditionalGeneration.from_pretrained(
     "Qwen/Qwen2-VL-2B-Instruct",
-    torch_dtype=torch.float16,
+    torch_dtype=torch.float32,
     trust_remote_code=True,
     device_map="auto",
 )
@@ -50,8 +49,8 @@ def post_process_text(text):
 def ocr(image):
     queries = [
         "Extract and transcribe all the text visible in the image, including any small or partially visible text.",
-        "Look closely at the image and list any text you see, no matter how small or unclear.",
-        "What text can you identify in this image? Include everything, even if it's partially obscured or in the background."
+        # "Look closely at the image and list any text you see, no matter how small or unclear.",
+        # "What text can you identify in this image? Include everything, even if it's partially obscured or in the background."
     ]
 
     all_extracted_text = []
@@ -71,6 +70,8 @@ def main_fun(image, keyword):
     
     if keyword:
         highlight_text = re.sub(f'({re.escape(keyword)})', r'<span style="background-color: yellow;">\1</span>', ext_text, flags=re.IGNORECASE)
+    else:
+        highlight_text = ext_text
     
     return ext_text, highlight_text