try the full model (no quantization)

main.py

@@ -63,8 +63,8 @@ async def lifespan(app: FastAPI):
         
         scripted_model_path = MODEL_DIR / SCRIPTED_MODEL_NAME
         metadata_model_path = MODEL_DIR / METADATA_MODEL_NAME
-        if DEVICE.type == "cpu":
-            scripted_model_path = MODEL_DIR / QUANTIZED_MODEL_NAME
+        # if DEVICE.type == "cpu":
+        #     scripted_model_path = MODEL_DIR / QUANTIZED_MODEL_NAME
         
         if  not scripted_model_path.exists():
             logger.error(f"Traced model not found at {scripted_model_path}")