Installation on University of Helsinki clusters (Ukko2 and Kale)

First you need to have access rights to the clusters. See instructions for who can get access rights to Kale or to Ukko2.

To install dpEmu on Kale or Ukko2 clusters, first establish a ssh connection to the cluster:

ssh ukko2.cs.helsinki.fi

Or:

ssh kale.grid.helsinki.fi

To install dpEmu without the ability of running all of the examples, execute the following commands in remote terminal:

module load Python/3.7.0-intel-2018b
export SCIKIT_LEARN_DATA=$TMPDIR

cd $WRKDIR
git clone https://github.com/dpEmu/dpEmu.git
cd dpEmu
python3 -m venv venv
source venv/bin/activate
pip install -U pip setuptools wheel --cache-dir $TMPDIR
pip install -r requirements/base.txt --cache-dir $TMPDIR
pip install -e "git+https://github.com/cocodataset/cocoapi.git#egg=pycocotools&subdirectory=PythonAPI" --cache-dir $TMPDIR
pip install -e . --cache-dir $TMPDIR

In order to run all of the examples, you’ll also need to execute the following command:

pip install -r requirements/examples.txt --cache-dir $TMPDIR

Object detection example and notebooks requirements

Further installation steps are needed to run the object detection example or notebooks. Execute the following commands after all of the above:

module load CUDA/10.0.130
module load cuDNN/7.5.0.56-CUDA-10.0.130

git clone https://github.com/dpEmu/Detectron.git libs/Detectron
./scripts/install_detectron.sh
git clone https://github.com/dpEmu/darknet.git libs/darknet
./scripts/install_darknet.sh

Instructions and examples for running jobs on Kale or Ukko2.