I have an autonomous vehicle Rosmaster X3 Plus with Jetson Orin NX 16 GB, I think it has Getpack 6. I use a container with ROS Noetic:3.0.4, so it has only Ubuntu 20 and Python 3.9.5. No Python 3.10. I would like to install PyTorch with CUDA to test LLM like Gemma 3n. Is there any way to do it?