Upload app.py with huggingface_hub

app.py

@@ -28,11 +28,12 @@ def load_model():
     """Load model once and cache it"""
     print("Loading model...")
 
-    # Prepare model loading kwargs
+    # Prepare model loading kwargs with disk offloading for limited memory
     model_kwargs = {
         "device_map": "auto",
         "trust_remote_code": True,
         "low_cpu_mem_usage": True,
+        "offload_folder": "offload",  # Enable disk offloading for HF Space
     }
 
     # Use 8-bit quantization if enabled (saves memory)