This repository includes utilities to build and run CUDA Docker images.
See samples/
for a list of Dockerfile examples.
Generate or build CUDA Dockerfiles given CUDA version(s).
$ usage: ./cudock (generate|build) [version...]
$ ./cudock generate 7.0 7.5
'cuda-7.0' successfully generated
'cuda-7.5' successfully generated
$ ls -R
.: cuda-7.0 cuda-7.5
./cuda-7.0: Dockerfile
./cuda-7.5: Dockerfile
$ ./cudock build 7.5
$ docker images
REPOSITORY TAG IMAGE ID CREATED VIRTUAL SIZE
cuda 7.5 5eba96294224 About an hour ago 2.146 GB
cuda latest 5eba96294224 About an hour ago 2.146 GB
Alternatively, one can build a CUDA image directly from this repository:
$ docker build -t "cuda:latest" -f "cuda-7.5" github.com/nvidia-docker
NVIDIA CUDA docker wrapper
This script is analogous to docker
, except that it will take care of setting up the NVIDIA host driver environment within Docker containers.
GPUs are exported through a list of comma-separated IDs using the environment variable GPU
.
Note thas the numbering is similar to the one reported by nvidia-smi
.
$ GPU=0,1 ./nvidia-docker <docker-options>