Octeract Engine

Octeract Engine is a massively parallel MINLP solver. It is written in ORL (Octeract Reformulation Language).

The engine contains 14 Octeract solvers for different types of mathematical structure, each with their own algorithms, to a grand total of 1031 high-performance algorithms.

Octeract Engine supports the Python, C++ and Julia programming languages alongside several modelling languages such as AMPL, PYOMO, JuMP, GAMS and Mosel.

Usage

Octeract Engine 3.1.0 can be loaded by module loading with the following command:

module load octeract-engine/3.1.0/binary

Interactive jobs

After connecting to Bessemer (see Establishing a SSH connection), Octeract Engine can be used interactively by starting an interactive session with srun --pty bash -i and then issuing the commands:

module load octeract-engine/3.1.0/binary
octeract-engine /usr/local/packages/live/noeb/octeract-engine/3.1.0/binary/examples/nl/ex2_1_1.nl -d ${PWD}

Batch jobs

Octeract Engine can be used in both SMP (single node only) and MPI parallel environments - on Bessemer no parallel environment needs specifiying.

Important

It is important that you use the -d $SLURM_SUBMIT_DIR argument to instruct Octeract Engine where to save the output file.

Octeract Engine will spawn a SLURM sub-task and SLURM will empty the $TMPDIR directory between tasks preventing any subsequent file move operation.

Example job:

!/bin/bash
#SBATCH -J octeract-8core-test
#SBATCH -o "%j".out
#SBATCH --mail-user a.person@sheffield.ac.uk
#SBATCH --mail-type=ALL
#SBATCH -t 0:05:0 # Request 5 mins run time
#SBATCH --ntasks-per-node=8
#SBATCH --mem=8000
​
module load octeract-engine/3.1.0/binary
octeract-engine /usr/local/packages/live/noeb/octeract-engine/3.1.0/binary/examples/nl/ex2_1_1.nl -n$SLURM_NTASKS -d $SLURM_SUBMIT_DIR

Using Octeract Engine with Pyomo:

Integrating the Octeract Engine with Pyomo is straightforward using our Python module.

By creating a specific Python environment for Octeract Engine and Pyomo you can help keep libraries and executables managed and available without polluting your base environment. This process, followed by running an example, is shown below:

Hint

You only need to create the conda environment and install Pyomo once. To use it for subsequent jobs you need only run the command: source activate octeract-engine-pyomo

module load octeract-engine/3.1.0/binary
module load Anaconda3/2019.07
conda create -n octeract-engine-pyomo python=3.7
source activate octeract-engine-pyomo #Make sure to use source activate, NOT conda activate.
pip install pyomo
pyomo --version #Check this version is supported.
python3 /usr/local/packages/live/noeb/octeract-engine/3.1.0/binary/examples/pyomo/pyomo_example.py

The above instructions have been adjusted from the following documentation provided by Octeract at: https://docs.octeract.com/htg1005-how_to_use_pyomo_with_octeract_engine

Installation notes

Octeract Engine 3.1.0 was a binary installation provided from the following link (https://download.octeract.com/octeract-engine-3.1.0-Linux-Centos7.tar.gz) and was installed using the script install_octeract-engine.sh

The software was tested by running the example batch job supplied above.