@brad007 it doesn’t work, because you are trying to run a container for JetPack 5 (L4T R35.x) on JetPack 4. Presuming that you have JetPack 4.6.1 or newer installed, you would run the L4T R32.7.x containers (those will have Python 3.6 though)
You would need to rebuilt PyTorch from source for Python 3.9. You can attempt to do that by following the Build from Source instructions from this topic. Just install the dev packages for the python version that you want, and call setup.py with the python version that you want (e.g. python3.9 setup.py bdist_wheel)
Seeing as PyTorch is normally built for Python 3.6 on Nano, you might run into some unforseen issues, but if you search that topic and others around the forum, others have been able to do it (for example, Install PyTorch with Python 3.8 on Jetpack 4.4.1)