VM Setup for Training

Using the OpenStack solution at LRZ it is no hassle recreating/adjusting our VMs. If there are any general questions for the setup, have a look at the Tutorial or simply ask me. To shorten the time for preparing a fresh instance following steps should be executed:

1. Create kaggle directory (on host machine)

mkdir .kaggle

2. Transfer your kaggle.json to the vm (execute on local host)

scp <path_to_kaggle.json> <username>@<ip>:~/.kaggle

3. Save following block to an .sh and transfer it to the VM using scp

#! /bin/bash
apt update
apt-get install python3
apt-get install python3-venv
apt-get install unzip

git clone https://github.com/energeeks/ashrae-energy-prediction.git
cd ashrae-energy-prediction

python3 -m venv .venv
source .venv/bin/activate
pip3 install -r requirements.txt

kaggle competitions download -c ashrae-energy-prediction
mkdir -p data/raw
unzip ashrae-energy-prediction.zip -d data/raw

make data
make features

4. Execute .sh and call training methods over make

5. If you are using a GPU for Lightgbm you also have to install following dependencies:

apt-get install cmake libboost-all-dev
pip3 uninstall lightgbm
pip3 install lightgbm --install-option=--gpu --install-option="--opencl-include-dir=/usr/local/cuda/include/" --install-option="--opencl-library=/usr/local/cuda/lib64/libOpenCL.so"