docker.md

Run from a Docker container

Python container

This container assumes python2 (and CUDA if needed) installed, and simply install all repo's dependencies and launch bash session.

1. Install docker.

2. Make sure docker daemon is running (sudo service docker start).

3. Build the project image

docker build . -t bm

4. Run the container (--rm option will automatically remove container when it exits)

nvidia-docker run --rm --name=bm-container -it bm bash

Ubuntu-CUDA container

This container builds CUDA, cuDNN and python before installing dependencies.

1. Install additionally nvidia-docker.

2. Build the project image

docker build . -t bm-cuda -f Dockerfile-gpu

4. Run the container

nvidia-docker run --rm --name=bm-cuda-container -it bm-cuda