Skip to content

Latest commit

 

History

History
49 lines (34 loc) · 1.89 KB

BUILD.md

File metadata and controls

49 lines (34 loc) · 1.89 KB

Building the containers

The main build script can be executed with:

bash build.sh

It builds Dockerfiles (found in the src folder) into container images in the following order:

Dockerfile Container Remarks
base.Dockerfile nvaitc/ai-lab:YY.MM-base Ubuntu+Conda+Jupyter+PyData
tf2.Dockerfile nvaitc/ai-lab:YY.MM-tf2 ++ TensorFlow 2.x
full.Dockerfile nvaitc/ai-lab:YY.MM ++ PyTorch, RAPIDS
vnc.Dockerfile nvaitc/ai-lab:YY.MM-vnc ++ VNC Desktop, VirtualGL

Adding additional packages into a container

You can use any of the containers as a base to build your own containers. The entire historical directory of containers can be accessed on Docker Hub.

Building your own container

As a simple example, to add the MXNet framework into container:

# create and build this Dockerfile

FROM nvaitc/ai-lab:20.03
LABEL maintainer="You <[email protected]>"

# you need to use root user for apt-get or make install
#USER root
#RUN apt-get update && apt-get install some-package

# use notebook user for pip/conda
USER $NB_UID
RUN pip install --no-cache-dir mxnet-cu101mkl

# always switch back to notebook user at the end
USER $NB_UID

After which, we can build the Dockerfile:

docker build . -f mxnet.Dockerfile -t myCustomMxNet:latest

Where -f points to the Dockerfile and -t defines the name of the output container.