Jupyter Notebook on L4T Pytorch Container

I am attempting to try some PyTorch development on a Jetson Nano, and want to use the L4T Pytorch Container. I’ve pulled it, and it doesn’t seem to have Jupyter notebook installed in it. I’ve attempted to pip install it in the container, but get a compile error.

Is there a way to use Jupyter Notebook with the PyTorch L4T container?

Hi @bradpr, if you try the l4t-ml container instead, it already has JupyterLab installed and it will start the Jupyter server automatically when you start the container. l4t-ml also has PyTorch installed (and a bunch of other ML stuff)

If you prefer to install JupyterLab inside l4t-pytorch instead, you can see the procedure that was followed to install it in l4t-ml here:

https://github.com/dusty-nv/jetson-containers/blob/1e10908a104494a883f6855d1e9947827f2a17bc/Dockerfile.ml#L153

Awesome - L4t-ml looks to have what I need already in it. Thanks for that!
One problem I just encountered with it, however, is that the ipywidgets module doesn’t render in the notebook. It only displays a text string describing what the widget should be rendering. For instance, I’m using the image collection widget code from the intro course classifier example. When I run it in the l4t-ml container’s instance of jupyter notebook, I only get
VBox(children=(HBox(children=(Image(value=b'\xff\xd8\xff\xe0\x00\x10JFIF\x00\x01\x01\x00\x00\x01\x00\x01\x00\

I’ve attempted to update & reinstall the ipywidgets module, but that has not helped.

Any advice on how to get this working?

I just saw this thread, and am trying it now:

Ah yes, sorry about that - I patched that in the Dockerfile, so you would need to rebuild the container using the jetson-containers repo, or manually patch it like in this comment.