Update app.py

app.py

@@ -31,20 +31,27 @@ pytesseract.pytesseract.tesseract_cmd = "/usr/bin/tesseract"
 # Set cache directory for Hugging Face
 os.environ["HF_HOME"] = "/app/cache"
 
-# Load Qwen2.5-VL-2B-Instruct model on CPU
-model_name = "Qwen/Qwen2.5-VL-2B-Instruct"
+# Get Hugging Face token from environment variable
+hf_token = os.getenv("HF_TOKEN")
+if not hf_token:
+    logger.error("HF_TOKEN environment variable not set")
+    raise HTTPException(status_code=500, detail="HF_TOKEN environment variable not set")
+
+# Load Qwen2-VL-2B-Instruct model on CPU
+model_name = "Qwen/Qwen2-VL-2B-Instruct"
 try:
     model = Qwen2VLForConditionalGeneration.from_pretrained(
         model_name,
         torch_dtype=torch.float16,
         device_map="auto",
-        low_cpu_mem_usage=True
+        low_cpu_mem_usage=True,
+        token=hf_token
     )
-    processor = AutoProcessor.from_pretrained(model_name)
-    logger.info("Qwen2.5-VL-2B-Instruct model loaded successfully")
+    processor = AutoProcessor.from_pretrained(model_name, token=hf_token)
+    logger.info("Qwen2-VL-2B-Instruct model loaded successfully")
 except Exception as e:
-    logger.error(f"Failed to load Qwen2.5-VL-2B-Instruct model: {str(e)}")
-    raise HTTPException(status_code=500, detail="Failed to load Qwen2.5-VL-2B-Instruct model")
+    logger.error(f"Failed to load Qwen2-VL-2B-Instruct model: {str(e)}")
+    raise HTTPException(status_code=500, detail="Failed to load Qwen2-VL-2B-Instruct model")
 
 # In-memory caches (1-hour TTL)
 raw_text_cache = cachetools.TTLCache(maxsize=100, ttl=3600)
@@ -98,7 +105,7 @@ async def process_pdf_page(img, page_idx):
         return ""
 
 async def process_with_qwen(filename: str, raw_text: str):
-    """Process raw text with Qwen2.5-VL-2B-Instruct to extract structured data."""
+    """Process raw text with Qwen2-VL-2B-Instruct to extract structured data."""
     start_time = time.time()
     logger.info(f"Starting Qwen processing for {filename}, {log_memory_usage()}")