Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

nvdashboard stopped working with nvcr.io/nvidia/pytorch:22.08-py3

See original GitHub issue

I have a docker file based on

FROM nvcr.io/nvidia/pytorch:22.08-py3
RUN pip3 install jupyterlab-nvdashboard jupyterlab-tensorboard-pro

Which used to work fine for a couple of month. I started rebuilding it in October and I am getting the pannels to show up but no gpus metric appear. also and 500 internal error as show below.

I also tested the latest

pip install --upgrade --pre jupyterlab-nvdashboard but no luck
using different base as FROM nvcr.io/nvidia/pytorch:22.09-py3 but still same error

Issue Analytics

State:
Created 10 months ago
Reactions:2
Comments:12 (2 by maintainers)

Top GitHub Comments

2reactions

AHarounicommented, Nov 17, 2022

Thanks for your response. using bokeh==2 broke jupyter lab completly for me. it din’t even want to start However, I managed to get it to work using bokeh==2.4 so this worked for me as

FROM nvcr.io/nvidia/pytorch:22.08-py3
RUN pip install jupyterlab-nvdashboard bokeh==2.4

1reaction

jacobtomlinsoncommented, Nov 18, 2022

Ah sorry I should’ve been more specific with my bokeh pin suggestion. Glad things are working now with 2.4.1.

I think the 500 errors you are seeing are related to #114 so I’m going to close out this issue now as resolved and open a new one to track bokeh 3 support specifically.