nvidia-anaconda-docker

Docker container for running JUPYTER Notebook within an ANACONDA nvironment, exploiting NVIDIA-DOCKER for GPUs The ANACONDA environment is also provided with:

1. tensorflow-gpu (latest version, 1.6.0 when buildt)
2. keras
3. tensorboard jupyter extension

Requirements

1. NVIDIA driver (cuda-drivers, not whole cuda!) for your GPU
2. nvidia-docker installed
3. DOCKER

Building Dockerfile

docker build -t nvidia-anaconda-docker <PATH_TO_DOCKERFILE_ROOT_PATH>

Running container

nvidia-docker run -v <HOST_NOTEBOOK_DIR>:/tmp -p 8888:8888 nvidia-anaconda-docker

From Dockerhub image

nvidia-docker run -v <HOST_NOTEBOOK_DIR>:/tmp -p 8888:8888 luke035/nvidia-anaconda-docker

Notes

Buildt image requires approximately 5GB of disk space.

Acknowledgments

1. Nvidia docker
2. Tensorflow GPU Docker
3. Anaconda
4. Jupyter tensorboard