KNITRO¶

Initial setup¶

from pyoptinterface import knitro

model = knitro.Model()

You need to follow the instructions in Getting Started to set up the optimizer correctly.

If you want to manage the license of KNITRO manually, you can create a knitro.Env object and pass it to the constructor of the knitro.Model object, otherwise a check of the license will be performed when initializing the knitro.Model object.

env = knitro.Env()
model = knitro.Model(env)

For users who want to release the license immediately after the optimization, you can call the close method of all models created and the knitro.Env object.

env = knitro.Env()
model = knitro.Model(env)
# do something with the model
model.close()
env.close()

The capability of `knitro.Model`¶

Supported constraints¶

Constraint	Supported
`model.add_linear_constraint`	✅
`model.add_quadratic_constraint`	✅
`model.add_second_order_cone_constraint`	✅
`model.add_exp_cone_constraint`	❌
`model.add_sos_constraint`	❌
`model.add_nl_constraint`	✅

Supported model attribute ¶

Attribute	Get	Set
Name	❌	❌
ObjectiveSense	✅	✅
DualStatus	❌	❌
PrimalStatus	✅	❌
RawStatusString	✅	❌
TerminationStatus	✅	❌
BarrierIterations	✅	❌
DualObjectiveValue	❌	❌
NodeCount	✅	❌
NumberOfThreads	✅	✅
ObjectiveBound	✅	❌
ObjectiveValue	✅	❌
RelativeGap	✅	❌
Silent	❌	✅
SimplexIterations	❌	❌
SolverName	✅	❌
SolverVersion	✅	❌
SolveTimeSec	✅	❌
TimeLimitSec	✅	✅

Supported variable attribute ¶

Attribute	Get	Set
Value	✅	❌
LowerBound	✅	✅
UpperBound	✅	✅
Domain	✅	✅
PrimalStart	❌	✅
Name	✅	✅
IISLowerBound	❌	❌
IISUpperBound	❌	❌
ReducedCost	✅	❌

Supported constraint attribute ¶

Attribute	Get	Set
Name	✅	✅
Primal	✅	❌
Dual	✅	❌

Solver-specific operations¶

Parameters¶

For solver-specific parameters, we provide get_raw_parameter and set_raw_parameter methods to get and set the parameters.

model = knitro.Model()

# Set the value of a parameter by name
model.set_raw_parameter("algorithm", 1)
model.set_raw_parameter("feastol", 1e-8)
model.set_raw_parameter("opttol", 1e-8)

# Set the value of a parameter by ID (using knitro.KN constants)
model.set_raw_parameter(knitro.KN.PARAM_MAXIT, 1000)

We also provide knitro.KN to contain common constants from the KNITRO C API.

# Using constants for parameter IDs
model.set_raw_parameter(knitro.KN.PARAM_FEASTOL, 1e-6)

# Algorithm selection
model.set_raw_parameter(knitro.KN.PARAM_NLP_ALGORITHM, knitro.KN.NLP_ALG_BAR_DIRECT)

# Output level
model.set_raw_parameter(knitro.KN.PARAM_OUTLEV, knitro.KN.OUTLEV_ITER)

# MIP parameters
model.set_raw_parameter(knitro.KN.PARAM_MIP_METHOD, knitro.KN.MIP_METHOD_BB)
model.set_raw_parameter(knitro.KN.PARAM_MIP_OPTGAPREL, 1e-4)

Variable and Constraint Properties¶

Common variable and constraint properties are provided through PyOptInterface dedicated methods:

Variable methods:

Bounds: set_variable_lb, get_variable_lb, set_variable_ub, get_variable_ub
Type and name: set_variable_name, get_variable_name, set_variable_domain
Starting point: set_variable_start
Solution values: get_value, get_variable_rc

Constraint methods:

Name: set_constraint_name, get_constraint_name
Solution values: get_constraint_primal, get_constraint_dual

Usage examples:

Variable properties:

# Bounds
model.set_variable_lb(variable, 0.0)
lb = model.get_variable_lb(variable)
model.set_variable_ub(variable, 10.0)
ub = model.get_variable_ub(variable)

# Type and name
model.set_variable_name(variable, "x")
name = model.get_variable_name(variable)

# Starting point
model.set_variable_start(variable, 1.0)

# Solution values
value = model.get_value(variable)
rc = model.get_variable_rc(variable)

Constraint properties:

# Name
model.set_constraint_name(constraint, "c1")
name = model.get_constraint_name(constraint)

# Solution values
primal = model.get_constraint_primal(constraint)
dual = model.get_constraint_dual(constraint)

Support for KNITRO callbacks¶

Unfortunately, KNITRO’s callback interface is not supported in PyOptInterface at the moment.