Pytorch detectes only one GPU in Jetson AGX Orin

1

I’m working to develop a face detection in Jetson AGX Orin using PyTorch but when running

import torch
print(torch.cuda.device_count())
only one GPU is detected.
The FPS using the GPU approximately matches with the CPU detection.

2

Jetson Orins use a single GPU as seen in these technical specs so I’m unsure what your expectations are.

3

I’m using Jetson AGX Orin not the nano orin

4

The link I’ve posted shows the technical specs for different Orins:

image
image1565×335 53.1 KB

5

sorry I have very recently started to use an edge device, does that means that it has only one GPU? what about the number of cores?
so if Pytorch detects one gpu then it’s utilizing the full potential of the GPU?

6

Yes, it has one GPU with multiple cores. Also yes, PyTorch will launch CUDA kernels to fully utilize the device.

7

Now it’s clear thank you