Torch on Jetson Orin Nano Jetpack 6.2.1

Robotics & Edge Computing Jetson Systems Jetson Orin Nano
, ,
1

I am facing issues while installing Torch and torchvision, torchaudio for my Jetson Orin Nano with the Jetpack 6.2.1 and cuda version is 12.6.68.
I have downloaded the python wheels from this website

and when I’m trying to install it, its getting built on the cpu instead of the gpu,
And in the end, I am still not able to go further because when I’m training the model, it immediately crashes because when i type device =0, it tries to look for the GPU, but since there is no GPU, it crashes.

3

*** Please note that this reply is generated by LLM automatically ***

Based on the provided context, it appears that you are facing issues with installing Torch, torchvision, and torchaudio on your Jetson Orin Nano with Jetpack 6.2.1 and CUDA version 12.6.68. You have downloaded the Python wheels from the website https://pypi.jetson-ai-lab.io/jp6/cu12.6.8/index and are trying to install them, but the installation is being built on the CPU instead of the GPU.

To resolve this issue, you can try the following steps:

  1. Verify CUDA installation: Ensure that CUDA is properly installed on your Jetson Orin Nano. You can check the CUDA version by running the command nvcc --version in the terminal. If CUDA is not installed, you can install it by following the instructions in the Jetpack documentation.
  2. Install CUDA Toolkit: Install the CUDA Toolkit on your Jetson Orin Nano. You can download the CUDA Toolkit from the NVIDIA website and follow the installation instructions.
  3. Install cuDNN: Install cuDNN, which is a library of GPU-accelerated primitives for deep neural networks. You can download cuDNN from the NVIDIA website and follow the installation instructions.
  4. Install Torch and torchvision: Install Torch and torchvision using the pip package manager. You can use the following commands to install Torch and torchvision:

pip3 install torch torchvision torchaudio --index-url https://pypi.jetson-ai-lab.io/jp6/cu12.6.8/index

   This command will install Torch, torchvision, and torchaudio with CUDA support.
5. **Verify GPU support**: After installing Torch and torchvision, verify that the GPU is being used by running the following command:
   ```python
import torch
print(torch.cuda.is_available())

If the output is True, it indicates that the GPU is being used.

In comparison to other Jetson models, the Jetson Orin Nano has a similar architecture to the Jetson AGX Orin, but with a smaller form factor and lower power consumption. The Jetson Orin Nano is designed for edge AI applications and is compatible with the same software stack as the Jetson AGX Orin.

In terms of CUDA support, the Jetson Orin Nano has CUDA 12.6.68, which is the same version as the Jetson AGX Orin. However, the Jetson Orin Nano has a smaller GPU with 128 CUDA cores, compared to the 256 CUDA cores on the Jetson AGX Orin.

Overall, the Jetson Orin Nano is a powerful edge AI platform that can run complex AI models, but it requires careful configuration and optimization to achieve optimal performance.

*** Please give the thumbs up if you get answers you like. Or provide feedback to help us improve the answer. ***

4

I have tried all this and still facing the same issue.
How do I go about it?

6

Hi,

Could you try the package below instead:

Thanks.