Docker with the SCIP Optimization Suite and PySCIPOpt. Solving optimization problem (0-1 Knapsack) with Python.
In this tutorial, we demonstrate how to solve a small-scale optimization problem (0-1 Knapsack) with Python, and explain how to build a docker container with SCIP and PySCIPOpt, to solve the optimization problem inside the container.
Why to choose SCIP
SCIP is currently one of the fastest non-commercial solvers for mixed integer programming (MIP) and mixed integer nonlinear programming (MINLP). It’s regularly updated releasing a new version several times a year. In addition, it provides an easy-to-use Python API (PySCIPOpt) to the SCIP optimization software. PySCIPOpt is implemented in Cython, what gets a good speedup when building an optimization problem in comparison with raw Python code.
How to build a docker with SCIP
To build a docker container with SCIP Optimization Suite and PySCIPOpt installed, you need to clone this [repository][1]. It contains only the root directory with the following files:
.
├── Dockerfile
│── docker-compose.yml
├── knapsack.py # demo example implemented small-scale knapsack problem
└── ...
To clone the repo, use following:
$ git clone https://github.com/viktorsapozhok/docker-scip.git
Before building a container, you need to download the latest version of the SCIP Optimization Suite, currently it’s 7.0.2. SCIP is distributed under the Academic License, and you can download it from the official website.
Note, that you need to download deb installer. Copy it to the root directory (where the Dockerfile is located).
Then to build a docker image, you can issue docker-compose build
from the root directory:
$ docker-compose build
When building process is over, you can check that a new image is now in the local image store.
$ docker images
REPOSITORY TAG IMAGE ID CREATED SIZE
scip v0.1 78791bbde634 14 hours ago 519MB
Now let’s take a look at the Dockerfile to understand in details what happens when you run the building process.
The first instruction specifies the base image on which we add new layers. Here we use the slim variant of the official Docker Python image.
FROM python:3.9-slim
As a next step, we install the basic GCC/g++ compilers and libraries included into build-essential
package. It’s needed to compile debian packages. Along with it, we install all the SCIP Optimization Suite dependencies.
RUN apt-get update \
&& DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
build-essential \
libgfortran4 \
libcliquer1 \
libopenblas-dev \
libgsl23 \
libtbb2 \
wget \
&& wget -O libboost.deb "http://archive.ubuntu.com/ubuntu/pool/main/b/boost1.65.1/libboost-program-options1.65.1_1.65.1+dfsg-0ubuntu5_amd64.deb" \
&& dpkg -i libboost.deb \
&& rm libboost.deb
Now everything is ready to install the SCIP Optimization Suite. We copy the debian package to docker, install it and remove the installer.
ADD SCIPOptSuite-7.0.2-Linux-ubuntu.deb /
RUN dpkg -i SCIPOptSuite-7.0.2-Linux-ubuntu.deb \
&& rm SCIPOptSuite-7.0.2-Linux-ubuntu.deb
As a last step, we create a new user, copy the demo script knapsack.py
to
user’s home directory, and install PySCIPOpt.
RUN groupadd --gid 1000 user \
&& useradd --uid 1000 --gid 1000 --create-home --shell /bin/bash user \
&& chown -R "1000:1000" /home/user
RUN mkdir /home/user/scripts
ADD knapsack.py /home/user/scripts
USER user
RUN pip install pyscipopt
To verify, that SCIP was installed correctly, we can issue scip --version
in the running container.
$ docker-compose up -d
$ docker exec -it scip scip --version
SCIP version 7.0.2 [precision: 8 byte] [memory: block] [mode: optimized] [LP solver: SoPlex 5.0.2] [GitHash: e9d280a398]
Copyright (C) 2002-2020 Konrad-Zuse-Zentrum fuer Informationstechnik Berlin (ZIB)
External codes:
SoPlex 5.0.2 Linear Programming Solver developed at Zuse Institute Berlin (soplex.zib.de) [GitHash: e24c304e]
CppAD 20180000.0 Algorithmic Differentiation of C++ algorithms developed by B. Bell (www.coin-or.org/CppAD)
ZLIB 1.2.11 General purpose compression library by J. Gailly and M. Adler (zlib.net)
GMP 6.1.2 GNU Multiple Precision Arithmetic Library developed by T. Granlund (gmplib.org)
ZIMPL 3.4.0 Zuse Institute Mathematical Programming Language developed by T. Koch (zimpl.zib.de)
PaPILO 1.0.2 parallel presolve for integer and linear optimization (https://github.com/lgottwald/PaPILO) [GitHash: 62d2842]
bliss 0.73p Computing Graph Automorphism Groups by T. Junttila and P. Kaski (http://www.tcs.hut.fi/Software/bliss/)
Ipopt 3.13.2 Interior Point Optimizer developed by A. Waechter et.al. (www.coin-or.org/Ipopt)
Compiler: gcc 7.5.0
Packing knapsack
To demonstrate how to use PySCIPOpt, we show how to solve a small-scale knapsack problem for the case of multiple knapsacks.
Let’s assume, that we have a collection of items with different weights and values, and we want to pack a subset of items into five knapsacks (bins), where each knapsack has a maximum capacity 100, so the total packed value is a maximum.
Define a simple container class to store item parameters and initialize 15 items.
class Item:
def __init__(self, index, weight, value):
self.index = index
self.weight = weight
self.value = value
items = [
Item(1, 48, 10), Item(2, 30, 30), Item(3, 42, 25), Item(4, 36, 50), Item(5, 36, 35),
Item(6, 48, 30), Item(7, 42, 15), Item(8, 42, 40), Item(9, 36, 30), Item(10, 24, 35),
Item(11, 30, 45), Item(12, 30, 10), Item(13, 42, 20), Item(14, 36, 30), Item(15, 36, 25)
]
Introduce bins (knapsacks) in the similar fashion.
class Bin:
def __init__(self, index, capacity):
self.index = index
self.capacity = capacity
bins = [Bin(1, 100), Bin(2, 100), Bin(3, 100), Bin(4, 100), Bin(5, 100)]
As a next step, we create a solver instance.
from pyscipopt import Model, quicksum
model = Model()
We introduce the binary variables x[i, j]
indicating that item i
is packed into bin j
.
x = dict()
for _item in items:
for _bin in bins:
x[_item.index, _bin.index] = model.addVar(vtype="B")
Now we add the constraints which prevent the situations when the same item is packed into multiple bins. It says that each item can be placed in at most one bin.
for _item in items:
model.addCons(quicksum(x[_item.index, _bin.index] for _bin in bins) <= 1)
The following constraints require that the total weight packed in each knapsack don’t exceed its maximum capacity.
for _bin in bins:
model.addCons(
quicksum(
_item.weight * x[_item.index, _bin.index] for _item in items
) <= _bin.capacity)
Finally, we define an objective function as a total value of the packed items and run the optimization.
model.setObjective(
quicksum(
_item.value * x[_item.index, _bin.index]
for _item in items for _bin in bins
),
sense="maximize")
model.optimize()
See knapsack.py for more details.
Running SCIP solver inside docker
We copy script knapsack.py
to /home/user/scripts
directory inside the container when building an image:
RUN mkdir /home/user/scripts
ADD knapsack.py /home/user/scripts
To launch the script, we start the container in the detached mode:
$ docker-compose up -d
To run the script inside the container, we use docker exec
command.
The optimization displays the following output:
$ docker exec -it scip python scripts/knapsack.py
...
Bin 1
Item 6: weight 48, value 30
Item 13: weight 42, value 20
Packed bin weight: 90
Packed bin value : 50
Bin 2
Item 3: weight 42, value 25
Item 8: weight 42, value 40
Packed bin weight: 84
Packed bin value : 65
Bin 3
Item 4: weight 36, value 50
Item 5: weight 36, value 35
Item 10: weight 24, value 35
Packed bin weight: 96
Packed bin value : 120
Bin 4
Item 2: weight 30, value 30
Item 11: weight 30, value 45
Item 14: weight 36, value 30
Packed bin weight: 96
Packed bin value : 105
Bin 5
Item 9: weight 36, value 30
Item 15: weight 36, value 25
Packed bin weight: 72
Packed bin value : 55
Total packed value: 395.0
To stop the running container, use down
command:
docker-compose down