Exposing Port 8888 on tf-torch-conda version

[mpasha10]

Hi,

I am very much interested in your docker. Would it be possible to expose 8888 and run Jupyter Lab by default when container is loaded? currently, this cannot be done and when I start it with in the container, the port is not exposed outside!

Many thanks

[matifali]

What do you get when you run?

docker run --gpus all --rm -it -h dockerdl -p 8888:8888 matifali/dockerdl:tf-torch-conda jupyter lab --no-browser --port 8888 --ServerApp.token='' --ip='*'

---

I intentionally did not add jupyter-lab as the CMD. You can build a custom image using matifali/dockerdl as base and update the CMD.

Something like this should work.

FROM matifali/dockerdl:tf-torch-conda
CMD ["jupyter", "lab", "--no-browser", "--port", "8888", "--ServerApp.token=''", "--ip='*'"]

Then build the image and run it with
docker run -p 8888:8888 username/iamge:tag

[mpasha10]

both returns the following:
/opt/nvidia/nvidia_entrypoint.sh: line 67: exec: jupyter: not found

[matifali]

Interesting.
Also try to update the ENTRYPOINT, as we are not updating it in Dockerfile and it defaults to the base image's entrypoint.

FROM matifali/dockerdl:tf-torch-conda
ENTRYPOINT ["jupyter", "lab", "--no-browser", "--port=8888", "--ip=0.0.0.0", "--allow-root"]