Build

NOTE: You can only use the GPU version on machines with GPUs because libcuda.so.1 will otherwise be inaccessible

docker build -t cannin/jupyter-keras-tensorflow-tools:tf-1.4.0-devel-py3 -f Dockerfile .
docker build -t cannin/jupyter-keras-tensorflow-tools-sshd:tf-1.4.0-devel-py3 -f Dockerfile_ssh .

sudo nvidia-docker build -t cannin/jupyter-keras-tensorflow-tools:tf-1.4.0-devel-gpu-py3 -f Dockerfile_gpu .
sudo nvidia-docker build -t cannin/jupyter-keras-tensorflow-tools-sshd:tf-1.4.0-devel-gpu-py3 -f Dockerfile_ssh .

Run

## Jupyter
docker rm -f keras; docker run --name keras -v $(pwd):/notebooks -p 8888:8888 -p 6006:6006 -t cannin/jupyter-keras-tensorflow-tools:tf-1.4.0-devel-py3

docker rm -f keras; docker run --name keras -v $(pwd):/notebooks -p 8888:8888 -p 6006:6006 -t cannin/jupyter-keras-tensorflow-tools:tf-1.4.0-devel-py3 jupyter lab --allow-root --no-browser

## Bash
docker rm -f keras; docker run --name keras -i -v $(pwd):/notebooks -p 8888:8888 -p 6006:6006 -t cannin/jupyter-keras-tensorflow-tools:tf-1.4.0-devel-py3 bash

## Interactive shell
docker exec -it keras bash

SSH

docker rm -f keras; docker run -d --name keras -p 23:22 -p 8888:8888 -p 6006:6006 -v $(pwd):/notebooks -w /notebooks -t cannin/jupyter-keras-tensorflow-tools-sshd:tf-1.4.0-devel-py3

docker rm -f keras; docker run --name keras -p 23:22 -p 8888:8888 -p 6006:6006 -v $(pwd):/notebooks -w /notebooks -it cannin/jupyter-keras-tensorflow-tools-sshd:tf-1.4.0-devel-py3 bash

docker rm -f keras; docker run --name keras -p 23:22 -p 8888:8888 -p 6006:6006 -v $(pwd):/notebooks -w /notebooks -it cannin/jupyter-keras-tensorflow-tools-sshd:tf-1.4.0-devel-py3 jupyter lab --allow-root --no-browser

# First time access may be slow
sudo nvidia-docker rm -f keras; sudo nvidia-docker run --name keras -p 23:22 -p 8888:8888 -p 6006:6006 -v $(pwd):/notebooks -w /notebooks -it cannin/jupyter-keras-tensorflow-tools-sshd:tf-1.4.0-devel-gpu-py3

# NOTE: No GPU
docker rm -f keras; docker run --name keras -p 3333:22 -p 9999:8888 -p 6666:6006 -v $(pwd):/notebooks -w /notebooks -it cannin/jupyter-keras-tensorflow-tools-sshd:tf-1.4.0-devel-py3

docker exec -it keras bash
ssh -p 23 root@localhost

AWS Instructions

Instructions for AWS instance setup instructions

CUDA Installation

See https://github.com/cannin/aws-cuda-docker-install

Basic GPU Test

cd /usr/local/cuda-8.0/samples/quasirandomGenerator
make clean; make; ./quasirandomGenerator