Asking about coral-tflite-delegate in raspberry

Hello. My name is Rio.

I deploy the pre-trained TF-lite model for QnA with NodeJS in Raspberry Pi using @tensorflow-models/qna - npm

However, the performance is not good. I decided to add the Coral USB Accelerator and installed the coral-tflite-delegate - npm

Further, I confused to integrate the coral-tflite-delegate in the QnA. Since, the QnA package will automatically load the model.

My question, how to use the coral-tflite-delegate for QnA with TF-lite? Thank you…