try

create_app.py

@@ -32,8 +32,15 @@ def load_models():
         QWEN_TOKENIZER = AutoTokenizer.from_pretrained(model_name, device='auto')
         QWEN_TOKENIZER.pad_token_id = QWEN_TOKENIZER.eos_token_id
         print("QWEN TOKENIZER LOADED")
-        QWEN_MODEL = AutoModelForCausalLM.from_pretrained(model_name).half()
-        QWEN_MODEL = QWEN_MODEL.to(device)
+        try:
+            QWEN_MODEL = AutoModelForCausalLM.from_pretrained(model_name).half()
+            QWEN_MODEL = QWEN_MODEL.to(device)
+            print("QWEN MODEL LOADED")  # Typo fixed from your original code ("LOADED" vs. "LOADED")
+        except Exception as e:
+            print(f"ERROR LOADING QWEN MODEL: {str(e)}")
+            raise  # Re-raise to see full traceback
+        # QWEN_MODEL = AutoModelForCausalLM.from_pretrained(model_name).half()
+        # QWEN_MODEL = QWEN_MODEL.to(device)
         print("QWEN MODEL LOADED")
         MODELS_LOADED = True
     print("LOAD ENDED")