Pyomo can't find gurobi solver

Question

I am trying to run a simple problem to test my setup:

#! /usr/bin/env python3

import pyomo.core as pyomo
from pyomo.opt import SolverFactory, IOptSolver

model = pyomo.ConcreteModel()
model.d = pyomo.Var(initialize=1, bounds=(0,2,))
model.g = pyomo.Var(initialize=1, bounds=(0,2,))
model.s = pyomo.Var(initialize=1, bounds=(0,2,))
model.b = pyomo.Var(initialize=1, bounds=(0,2,))

objective_rule = model.d - model.b*model.g + model.s*model.b
model.objective = pyomo.Objective(rule=objective_rule, sense=pyomo.minimize)

# After digging in pyomo source I figured out this
# should print all solvers available in my system
print(IOptSolver._factory_cls)

# This fails
opt = SolverFactory("gurobi", solver_io="python")
# opt = SolverFactory("neos") This also fails

# Create a model instance and optimize
instance = model
results = opt.solve(instance)
instance.display()

But pyomo has problems detecting gurobi:

{}
Traceback (most recent call last):
  File "./__test__2.py", line 21, in <module>
    results = opt.solve(instance)
  File "/usr/local/lib/python3.5/dist-packages/pyomo/opt/base/solvers.py", line 125, in solve
    self._solver_error('solve')
  File "/usr/local/lib/python3.5/dist-packages/pyomo/opt/base/solvers.py", line 153, in _solver_error
    + "\n\toptions: %s" % ( self.options, ) )
RuntimeError: Attempting to use an unavailable solver.

The SolverFactory was unable to create the solver "gurobi"
and returned an UnknownSolver object.  This error is raised at the point
where the UnknownSolver object was used as if it were valid (by calling
method "solve").

The original solver was created with the following parameters:
    executable: gurobi
    solver_io: python
    type: gurobi
    _args: ()
    options: {}

I've installed pyomo using pip:

pip3 install pyomo

And I've downloaded and installed gurobi using the following commands (after decompress):

mv gurobi801 /opt/gurobi/gurobi801
cd /opt/gurobi/gurobi801/linux64/
python3 setup.py build
python3 setup.py install

And added to my .bash_profile:

export GUROBI_HOME="/opt/gurobi/gurobi801/linux64"
export PATH="${PATH}:${GUROBI_HOME}/bin"

Gurobi runs fine if I call it from command line:

gurobi.sh 
Python 2.7.13 (default, Sep  4 2017, 15:40:17) 
[GCC 4.4.7 20120313 (Red Hat 4.4.7-18)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
Academic license - for non-commercial use only

Gurobi Interactive Shell (linux64), Version 8.0.1
Copyright (c) 2018, Gurobi Optimization, LLC
Type "help()" for help

gurobi>

But pyomo fails to print the whole help:

pyomo help -s

Pyomo Solvers and Solver Managers
---------------------------------
Pyomo uses 'solver managers' to execute 'solvers' that perform
optimization and other forms of model analysis.  A solver directly
executes an optimizer, typically using an executable found on the
user's PATH environment.  Solver managers support a flexible mechanism
for asyncronously executing solvers either locally or remotely.  The
following solver managers are available in Pyomo:

    neos       Asynchronously execute solvers on the NEOS server
    serial     Synchronously execute solvers locally

If no solver manager is specified, Pyomo uses the serial solver
manager to execute solvers locally.  The pyro and phpyro solver
managers require the installation and configuration of the pyro
software.  The neos solver manager is used to execute solvers on the
NEOS optimization server.


Serial Solver Interfaces
------------------------
The serial, pyro and phpyro solver managers support the following
solver interfaces:

Traceback (most recent call last):
  File "/usr/local/lib/python3.5/dist-packages/pyomo/scripting/driver_help.py", line 338, in help_solvers
    logger.disable(logging.WARNING)
AttributeError: 'Logger' object has no attribute 'disable'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/usr/local/bin/pyomo", line 11, in <module>
    sys.exit(main())
  File "/usr/local/lib/python3.5/dist-packages/pyomo/scripting/pyomo_main.py", line 82, in main
    retval = _options.func(_options)
  File "/usr/local/lib/python3.5/dist-packages/pyomo/scripting/driver_help.py", line 452, in help_exec
    help_solvers()
  File "/usr/local/lib/python3.5/dist-packages/pyomo/scripting/driver_help.py", line 351, in help_solvers
    logger.disable(logging.NOTSET)
AttributeError: 'Logger' object has no attribute 'disable'

What could I have missed in my setup? I'm running ubuntu 16.04 x64

Answer 1

It turns out importing pyomo.environ is of vital importance to detect the available solvers. Updating the script to:

#! /usr/bin/env python3

import pyomo.environ # <---  HAD MISSING #
from pyomo.opt import SolverFactory, IOptSolver

import pyomo.core as pyomo

model = pyomo.ConcreteModel()
model.d = pyomo.Var(initialize=0, bounds=(0,2,))
model.g = pyomo.Var(initialize=0, bounds=(0,2,))
model.s = pyomo.Var(initialize=0, bounds=(0,2,))
model.b = pyomo.Var(initialize=0, bounds=(0,2,))

objective_rule = model.d - model.b*model.g + model.s*model.b
model.objective = pyomo.Objective(rule=objective_rule, sense=pyomo.minimize)

# Should print all available solvers
print(sorted(IOptSolver._factory_cls.keys()))
opt = SolverFactory("gurobi", solver_io="python")

# Create a model instance and optimize
instance = model
results = opt.solve(instance)
instance.display()

Yields:

['_cbc_shell', '_cplex_shell', '_gams_direct', '_gams_shell', '_glpk_direct', '_glpk_shell', '_glpk_shell_4_42', '_glpk_shell_old', '_gurobi_shell', '_mock_asl', '_mock_cbc', '_mock_cplex', '_mock_glpk', '_mock_pico', '_mock_xpress', '_neos', '_pico_shell', '_xpress_shell', 'asl', 'baron', 'bilevel_blp_global', 'bilevel_blp_local', 'bilevel_ld', 'cbc', 'conopt', 'cplex', 'cplex_direct', 'cplex_persistent', 'gams', 'gdpopt', 'glpk', 'gurobi', 'gurobi_direct', 'gurobi_persistent', 'ipopt', 'mpec_minlp', 'mpec_nlp', 'path', 'pico', 'ps', 'py', 'scip', 'trustregion', 'xpress']
Academic license - for non-commercial use only
Model unknown

  Variables:
    d : Size=1, Index=None
        Key  : Lower : Value : Upper : Fixed : Stale : Domain
        None :     0 :     0 :     2 : False :  True :  Reals
    g : Size=1, Index=None
        Key  : Lower : Value : Upper : Fixed : Stale : Domain
        None :     0 :     0 :     2 : False :  True :  Reals
    s : Size=1, Index=None
        Key  : Lower : Value : Upper : Fixed : Stale : Domain
        None :     0 :     0 :     2 : False :  True :  Reals
    b : Size=1, Index=None
        Key  : Lower : Value : Upper : Fixed : Stale : Domain
        None :     0 :     0 :     2 : False :  True :  Reals

  Objectives:
    objective : Size=1, Index=None, Active=True
        Key  : Active : Value
        None :   True :   0.0

  Constraints:
    None

Is not doing proper optimization but at least the solver detection issue is solved.