README.md

Docker Builds

If you've never used Docker before, check out the getting started guide.

Up-to-date prebuilt images can be found on Docker hub.

CPU Images

• cpu-latest, cpu, cpu-0.1, latest
• cpu-jupyter-kubeflow-latest, cpu-jupyter-kubeflow, cpu-jupyter-kubeflow-0.1

GPU Images

• gpu-latest, gpu, gpu-0.1
• gpu-jupyter-kubeflow-latest, gpu-jupyter-kubeflow, gpu-jupyter-kubeflow-0.1

The CPU images should run out of the box without any driver prerequisites.

GPU Images

Prerequisites

To use the GPU images, you need to have:

• The Nvidia drivers
• The Nvidia-docker container runtime
• Docker configured to use the Nvidia container runtime

Once you have those installed, you should be ready to start running the GPU instances.

Testing GPUs

To make sure your GPU instance is setup correctly, run the following in a container:

import dlib
print(dlib.cuda.get_num_devices())

Jupyter Images

The Jupyter images are built to be deployed on Kubeflow. However, if you just want to run a normal Jupyter instance, they're a great template to build your own.

Example Dockerfile

Here's an example Dockerfile using the prebuilt images:

FROM animcogn/face_recognition:gpu

COPY requirements.txt requirements.txt

RUN pip3 install -r ./requirements.txt

COPY my_app /my_app

CMD [ "python3", "/my_app/my_app.py" ]