How to tell Core ML to use eGPU?

I am using this piece of software (StarXTerminator) for removing stars in an astrophoto. This program uses Core ML heavily in macOS. Core ML would use GPU to accelerate the process.

If I run StarXTerminator on 2019 Mac Pro (macOS 12.6) equipped with AMD 5700XT card installed, it would use 5700XT flawlessly. However, if I use StarXTerminator on 2018 Macbook Pro (macOS 12.6) with 5700XT sitting in Akitio eGPU box, it always chose to ignore eGPU and used Radeon Pro 560X inside Macbook Pro.

The author of StarXTerminator said he just throw the data to Core ML and let it decide whatever GPU was available for the task. It's out of his hands on which GPU to use.

The question: is there any way to force Core ML library to use eGPU on Macbook Pro? Terminal commands? Utilities?

Thanks for any telp.

Hojong Lin