• Pytorch: Docker4pt
• Tensorflow: Docker4tensorflow_gpu

For each environment cd to the directory, then do the build and run. For example, for the Pytorch environment:

cd Docker4pt
bash build.sh

# Run a container for the 1st time
bash docker_run.sh run

# Get into a built container
bash docker_run.sh bash
# OR
docker exec -it CONTAINER_NAME bash

# To remove a container
bash docker_run.sh rm

To run jupyter notebook and tensorboard: Ports 8888 and 6006 are default ports for running Jupyter notebook and tensorboard.

• Jupyter-notebook:

jupyter-notebook --allow-root --port 8888 --ip=0.0.0.0 --no-browser

• Jupyter-lab (Recommended):

jupyter-lab --allow-root --port 8888 --ip=0.0.0.0 --no-browser

For Jupyter-notebook and Jupyter-lab, multiple people could use the same session, simply run the following to get the token for the token.

jupyter server list

• Tensorboard:

tensorboard --logdir <dir to keep the logs> --host 0.0.0.0 --port 6006

• You may add additional packages to the DockerFile
• For permission purposes, you need to add your uid to the Dockerfile as follows (note to use your own netid and uid):

RUN echo "zyang37:x:3944:" >> /etc/group
RUN echo "zyang37:x:26388:3944:Zhenning Yang,620,,:/home/zyang37:/bin/bash" >> /etc/passwd