Python AI notebook stack

---

Python 3 + Jupyter Notebook (with themes) + TensorFlow + PyTorch + TensorBoard + CUDA/CuDNN

Excellent for AI research on GPU(s) in Python. The Dockerfile is heavily based on TensorFlow dockerfiles.

Prerequisites

Requires Docker (>=19.03), NVIDIA driver and nvidia-docker (see how to install it here). Tested on Debian 10, should work on Ubuntu, CentOS and RHEL as well.

Usage

Use the following command to run the notebook:

docker-compose up 

If you want to make your adjustments (like I do with the KGTN repo):

docker-compose -f docker-compose.yml -f docker-compose-KGTN.yml up 

Now navigate to http://127.0.0.1:18888. The default password is: Cybermyszki7

The following code will fire up tensorboard:

docker exec -it [container_name] tensorboard --logdir=/tf/logs --bind_all

Now you can navigate to http://127.0.0.1:16006 to access tensorboard.

Additional information

You can change the default password. In Python (pip install ipython if needed) do the following:

from IPython.lib import passwd
passwd()
> Enter password:
> Verify password:
> 'sha1:xxxxxxx:xxxxxxxxxxxxxxx'

Now you have to copy the SHA hash to the Dockerfile and built it on your computer (might take 15 minutes or so).

docker build -t 0xdfdfdf/ai-notebook .