Using Jetson Nano as an Auxiliary GPU

Autonomous Machines Jetson & Embedded Systems Jetson Nano
1

Hi guys, sorry anything, I’m new to Jetson.

I’m working on a project with learning by reinforcement, robotic simulation. This project is centered on a Windows notebook, without GPU.

I would like to use my Jetson Nano as an auxiliary GPU, harnessing the power of the CUDA so that python can transfer training directly to the external Jetson nano. It would be similar to a communication process. For this to happen I believe that Python (tensorflow) should recognize the Jetson Nano as my GPU.

Is there any way to make this work, ie using the Jetson Nano as an auxiliary GPU?

Thanks.

2

There are ways to communicate between Nano and PC, but appearing as a GPU is not one of those possibilities. The Nano is not particularly good at training AI, but is mostly intended to use pre-trained models. You could in theory train with the Nano, but I wouldn’t recommend it. You can run a Python program which is able to talk over a network to other computers and it won’t care what type of computer it is at the other end. What you won’t find is that such a program is suitable to look like a GPU, although it might work well to transfer data produced by a Jetson’s GPU.