imyutaro / dockerfiles

Dockerfiles

Dockerfiles for deep learning experiments.

All Dockerfiles below include

• Python
• Jupyter lab

CPU ver

Using docker-compose.

• Dockerfile-python
  • Base image is python3.
• docker-compose.yml
  • Set options for docker-compose up(docker run).
• start.sh
  • Shell script for docker-compose up.

# Build and run docker container
./start.sh Dockerfile_name
# Enter a docker container
docker exec -it experiment zsh

# If you want to use specific container and image name
NAME=cpu ./start.sh
# If you want to use specific port for jupyter lab (default port 8888)
PORT=9000 ./start.sh

GPU ver

I want to use docker-compose but docker-compose doesn't support gpus option, so I use docker build and docker run.

• Dockerfile-nvidia
  • Base image is nvidia.
• Dockerfile-pytorch
  • Base image is pytorch.
• start_gpu.sh
  • Shell script for docker build and docker run.

# Build and run docker container
./start_gpu.sh Dockerfile_name
# Enter a docker container
docker exec -it experiment-gpu zsh

# If you want to use specific container and image name
NAME=gpu ./start.sh
# If you want to use specific port for jupyter lab (default port 8888)
PORT=9000 ./start.sh