Using LightGBM via Docker

This directory contains Dockerfiles to make it easy to build and run LightGBM via Docker.

Installing Docker

Follow the general installation instructions on the Docker site:

• macOS
• Ubuntu
• Windows

Using CLI Version of LightGBM via Docker

Build a Docker image with LightGBM CLI:

mkdir lightgbm-docker
cd lightgbm-docker
wget https://raw.githubusercontent.com/Microsoft/LightGBM/master/docker/dockerfile-cli
docker build -t lightgbm-cli -f dockerfile-cli .

where lightgbm-cli is the desired Docker image name.

Run the CLI from the container:

docker run --rm -it \
--volume $HOME/lgbm.conf:/lgbm.conf \
--volume $HOME/model.txt:/model.txt \
--volume $HOME/tmp:/out \
lightgbm-cli \
config=lgbm.conf

In the above example, three volumes are mounted from the host machine to the Docker container:

• lgbm.conf - task config, for example

app=multiclass
num_class=3
task=convert_model
input_model=model.txt
convert_model=/out/predict.cpp
convert_model_language=cpp

• model.txt - an input file for the task, could be training data or, in this case, a pre-trained model.
• out - a directory to store the output of the task, notice that convert_model in the task config is using it.

config=lgbm.conf is a command-line argument passed to the lightgbm executable, more arguments can be passed if required.

Running the Python-package Сontainer

Build the container, for Python users:

mkdir lightgbm-docker
cd lightgbm-docker
wget https://raw.githubusercontent.com/Microsoft/LightGBM/master/docker/dockerfile-python
docker build -t lightgbm -f dockerfile-python .

After build finished, run the container:

docker run --rm -it lightgbm

Running the R-package Сontainer

Build the container based on the verse Rocker image, for R users:

mkdir lightgbm-docker
cd lightgbm-docker
wget https://raw.githubusercontent.com/Microsoft/LightGBM/master/docker/dockerfile-r
docker build -t lightgbm-r -f dockerfile-r .

This will default to the latest version of R. If you want to try with an older rocker container to run a particular version of R, pass in a build arg with a valid tag.

For example, to test with R 3.5:

docker build \
    -t lightgbm-r-35 \
    -f dockerfile-r \
    --build-arg R_VERSION=3.5 \
    .

After the build is finished you have two options to run the container:

1. Start RStudio, an interactive development environment, so that you can develop your analysis using LightGBM or simply try out the R package. You can open RStudio in your web browser.
2. Start a regular R session.

In both cases you can simply call

library("lightgbm")

to load the installed LightGBM R package.

RStudio

docker run --rm -it -e PASSWORD=lightgbm -p 8787:8787 lightgbm-r

Open the browser at http://localhost:8787 and log in. See the rocker/rstudio image documentation for further configuration options.

Regular R

If you just want a vanilla R process, change the executable of the container:

docker run --rm -it lightgbm-r R