petsc4py.PETSc.TAO#
- class petsc4py.PETSc.TAO#
Bases:
ObjectOptimization solver.
TAO is described in the
PETSc manual.See also
Enumerations
TAO Augmented Lagrangian Multiplier method (ALMM) Type.
TAO Bound Constrained Conjugate Gradient (BNCG) Update Type.
TAO solver termination reason.
TAO solver type.
Methods Summary
appendOptionsPrefix(prefix)Append to the prefix used for searching for options in the database.
Cancel all the monitors of the solver.
computeConstraints(x, c)Compute the vector corresponding to the constraints.
computeDualVariables(xl, xu)Compute the dual vectors corresponding to variables' bounds.
computeGradient(x, g)Compute the gradient of the objective function.
computeHessian(x, H[, P])Compute the Hessian of the objective function.
computeJacobian(x, J[, P])Compute the Jacobian.
Compute the value of the objective function.
computeObjectiveGradient(x, g)Compute the gradient of the objective function and its value.
computeResidual(x, f)Compute the residual.
computeVariableBounds(xl, xu)Compute the vectors corresponding to variables' bounds.
create([comm])Create a TAO solver.
createPython([context, comm])Create an optimization solver of Python type.
destroy()Destroy the solver.
Return the subsolver inside the ALMM solver.
Return the type of the ALMM solver.
Return the application context.
Return the type of the BNCG solver.
Return the damping vector.
Return the subsolver inside the BRGN solver.
Return the constraints tolerance parameters used in the convergence tests.
Return the reason for the solver convergence.
Return the callback used to test for solver convergence.
Return tuple holding vector and callback of equality constraints.
Return the vector used to store the gradient and the evaluation callback.
Return the matrix used to compute inner products.
Return the matrices used to store the Hessian and the evaluation callback.
Return tuple holding vector and callback of inequality constraints.
Return the current iteration number.
Return matrix, precon matrix and callback of equality constraints Jacobian.
Return matrix, precon matrix and callback of ineq.
getKSP()Return the linear solver used by the nonlinear solver.
Return the initial Hessian for the quasi-Newton approximation.
Return the
KSPfor the inverse of the initial Hessian approximation.Get the LMVM matrix.
Return the TAO Line Search object.
Return the maximum number of objective evaluations within the solver.
Return the maximum number of solver iterations.
Return the callback used to monitor solver convergence.
Return the objective evaluation callback.
Return the vector used to store the gradient and the evaluation callback.
Return the current value of the objective function.
Return the prefix used for searching for options in the database.
Return the instance of the class implementing the required Python methods.
Return the fully qualified Python name of the class used by the solver.
Return the vector holding the solution.
Return the objective function value and the norms of gradient and constraints.
Return the solution status.
Return the tolerance parameters used in the solver convergence tests.
getType()Return the type of the solver.
Return the callback to compute the update.
Return the lower and upper bounds vectors.
monitor([its, f, res, cnorm, step])Monitor the solver.
setALMMSubsolver(subsolver)Set the subsolver inside the ALMM solver.
setALMMType(tao_almm_type)Set the ALMM type of the solver.
setAppCtx(appctx)Set the application context.
setBNCGType(cg_type)Set the type of the BNCG solver.
Set the dictionary matrix.
setBRGNRegularizerHessian(hessian[, H, ...])Set the callback to compute the regularizer Hessian.
setBRGNRegularizerObjectiveGradient(objgrad)Set the callback to compute the regularizer objective and gradient.
setBRGNRegularizerWeight(weight)Set the regularizer weight.
setBRGNSmoothL1Epsilon(epsilon)Set the smooth L1 epsilon.
setConstraintTolerances([catol, crtol])Set the constraints tolerance parameters used in the solver convergence tests.
setConstraints(constraints[, C, args, kargs])Set the callback to compute constraints.
setConvergedReason(reason)Set the termination flag.
setConvergenceTest(converged[, args, kargs])Set the callback used to test for solver convergence.
setEqualityConstraints(equality_constraints, c)Set equality constraints callback.
Configure the solver from the options database.
setGradient(gradient[, g, args, kargs])Set the gradient evaluation callback.
setGradientNorm(mat)Set the matrix used to compute inner products.
setHessian(hessian[, H, P, args, kargs])Set the callback to compute the Hessian matrix.
setInequalityConstraints(...[, args, kargs])Set inequality constraints callback.
setInitialTrustRegionRadius(radius)Set the initial trust region radius.
setIterationNumber(its)Set the current iteration number.
setJacobian(jacobian[, J, P, args, kargs])Set the callback to compute the Jacobian.
setJacobianDesign(jacobian_design[, J, ...])Set Jacobian design callback.
setJacobianEquality(jacobian_equality[, J, ...])Set Jacobian equality constraints callback.
setJacobianInequality(jacobian_inequality[, ...])Set Jacobian inequality constraints callback.
setJacobianResidual(jacobian[, J, P, args, ...])Set the callback to compute the least-squares residual Jacobian.
setJacobianState(jacobian_state[, J, P, I, ...])Set Jacobian state callback.
setLMVMH0(mat)Set the initial Hessian for the quasi-Newton approximation.
setLMVMMat(M)Set the LMVM matrix.
Set the maximum number of objective evaluations within the solver.
setMaximumIterations(mit)Set the maximum number of solver iterations.
setMonitor(monitor[, args, kargs])Set the callback used to monitor solver convergence.
setObjective(objective[, args, kargs])Set the objective function evaluation callback.
setObjectiveGradient(objgrad[, g, args, kargs])Set the objective function and gradient evaluation callback.
setOptionsPrefix(prefix)Set the prefix used for searching for options in the database.
setPythonContext(context)Set the instance of the class implementing the required Python methods.
setPythonType(py_type)Set the fully qualified Python name of the class to be used.
setResidual(residual, R[, args, kargs])Set the residual evaluation callback for least-squares applications.
setSolution(x)Set the vector used to store the solution.
setStateDesignIS([state, design])Set the index sets indicating state and design variables.
setTolerances([gatol, grtol, gttol])Set the tolerance parameters used in the solver convergence tests.
setType(tao_type)Set the type of the solver.
setUp()Set up the internal data structures for using the solver.
setUpdate(update[, args, kargs])Set the callback to compute update at each optimization step.
setVariableBounds(varbounds[, args, kargs])Set the lower and upper bounds for the optimization problem.
solve([x])Solve the optimization problem.
view([viewer])View the solver.
Attributes Summary
Application context.
Constraints norm.
Boolean indicating if the solver has converged.
Broken.
Boolean indicating if the solver has failed.
Broken.
Objective value.
Gradient norm.
Gradient vector.
Broken.
Boolean indicating if the solver has not converged yet.
Number of iterations.
Linear solver.
Objective value.
Converged reason.
Solution vector.
Methods Documentation
- appendOptionsPrefix(prefix)#
Append to the prefix used for searching for options in the database.
Logically collective.
- cancelMonitor()#
Cancel all the monitors of the solver.
Logically collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:1373
- Return type:
- computeConstraints(x, c)#
Compute the vector corresponding to the constraints.
Collective.
- Parameters:
- Return type:
See also
- computeDualVariables(xl, xu)#
Compute the dual vectors corresponding to variables’ bounds.
Collective.
See also
- computeGradient(x, g)#
Compute the gradient of the objective function.
Collective.
See also
- computeHessian(x, H, P=None)#
Compute the Hessian of the objective function.
Collective.
- Parameters:
- Return type:
See also
- computeJacobian(x, J, P=None)#
Compute the Jacobian.
Collective.
- Parameters:
- Return type:
See also
- computeObjective(x)#
Compute the value of the objective function.
Collective.
See also
- computeObjectiveGradient(x, g)#
Compute the gradient of the objective function and its value.
Collective.
- Parameters:
- Return type:
- computeResidual(x, f)#
Compute the residual.
Collective.
See also
- computeVariableBounds(xl, xu)#
Compute the vectors corresponding to variables’ bounds.
Collective.
See also
- create(comm=None)#
Create a TAO solver.
Collective.
- Parameters:
comm (Comm | None) – MPI communicator, defaults to
Sys.getDefaultComm.- Return type:
See also
- createPython(context=None, comm=None)#
Create an optimization solver of Python type.
Collective.
- Parameters:
context (Any) – An instance of the Python class implementing the required methods.
comm (Comm | None) – MPI communicator, defaults to
Sys.getDefaultComm.
- Return type:
- destroy()#
Destroy the solver.
Collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:139
- Return type:
- getALMMSubsolver()#
Return the subsolver inside the ALMM solver.
Not collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:1717
- Return type:
- getALMMType()#
Return the type of the ALMM solver.
Not collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:1732
- Return type:
- getAppCtx()#
Return the application context.
Source code at petsc4py/PETSc/TAO.pyx:293
- Return type:
- getBNCGType()#
Return the type of the BNCG solver.
Not collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:1569
- Return type:
- getBRGNDampingVector()#
Return the damping vector.
Not collective.
Source code at petsc4py/PETSc/TAO.pyx:1862
- Return type:
- getBRGNSubsolver()#
Return the subsolver inside the BRGN solver.
Not collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:1779
- Return type:
- getConstraintTolerances()#
Return the constraints tolerance parameters used in the convergence tests.
Not collective.
- Returns:
- Return type:
- getConvergedReason()#
Return the reason for the solver convergence.
Not collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:1626
- Return type:
- getConvergenceTest()#
Return the callback used to test for solver convergence.
Not collective.
See also
- getEqualityConstraints()#
Return tuple holding vector and callback of equality constraints.
Not collective.
- getGradient()#
Return the vector used to store the gradient and the evaluation callback.
Not collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:436
- Return type:
- getGradientNorm()#
Return the matrix used to compute inner products.
Not collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:1484
- Return type:
- getHessian()#
Return the matrices used to store the Hessian and the evaluation callback.
Not collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:598
- Return type:
- getInequalityConstraints()#
Return tuple holding vector and callback of inequality constraints.
Not collective.
- getIterationNumber()#
Return the current iteration number.
Not collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:1596
- Return type:
- getJacobianEquality()#
Return matrix, precon matrix and callback of equality constraints Jacobian.
Not collective.
- getJacobianInequality()#
Return matrix, precon matrix and callback of ineq. constraints Jacobian.
Not collective.
- getKSP()#
Return the linear solver used by the nonlinear solver.
Not collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:1700
- Return type:
- getLMVMH0()#
Return the initial Hessian for the quasi-Newton approximation.
Not collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:1511
- Return type:
- getLMVMH0KSP()#
Return the
KSPfor the inverse of the initial Hessian approximation.Not collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:1526
- Return type:
- getLMVMMat()#
Get the LMVM matrix.
Not collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:709
- Return type:
- getLineSearch()#
Return the TAO Line Search object.
Not collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:1959
- Return type:
- getMaximumFunctionEvaluations()#
Return the maximum number of objective evaluations within the solver.
Not collective.
Source code at petsc4py/PETSc/TAO.pyx:1204
- Return type:
- getMaximumIterations()#
Return the maximum number of solver iterations.
Not collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:1177
- Return type:
- getMonitor()#
Return the callback used to monitor solver convergence.
Not collective.
See also
- getObjective()#
Return the objective evaluation callback.
Not collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:423
- Return type:
- getObjectiveAndGradient()#
Return the vector used to store the gradient and the evaluation callback.
Not collective.
Source code at petsc4py/PETSc/TAO.pyx:482
- Return type:
- getObjectiveValue()#
Return the current value of the objective function.
Not collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:1610
- Return type:
- getOptionsPrefix()#
Return the prefix used for searching for options in the database.
Not collective.
Source code at petsc4py/PETSc/TAO.pyx:234
- Return type:
- getPythonContext()#
Return the instance of the class implementing the required Python methods.
Not collective.
Source code at petsc4py/PETSc/TAO.pyx:1914
- Return type:
- getPythonType()#
Return the fully qualified Python name of the class used by the solver.
Not collective.
See also
PETSc Python optimization solver type (TODO),
setPythonContext,setPythonType,TaoPythonGetTypeSource code at petsc4py/PETSc/TAO.pyx:1944
- Return type:
- getSolution()#
Return the vector holding the solution.
Not collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:1457
- Return type:
- getSolutionNorm()#
Return the objective function value and the norms of gradient and constraints.
Not collective.
- Returns:
- Return type:
See also
- getSolutionStatus()#
Return the solution status.
Not collective.
- Returns:
its (
int) – Current number of iterations.f (
float) – Current value of the objective function.res (
float) – Current value of the residual norm.cnorm (
float) – Current value of the constrains norm.step (
float) – Current value of the step.reason (
ConvergedReason) – Current value of converged reason.
- Return type:
See also
- getTolerances()#
Return the tolerance parameters used in the solver convergence tests.
Not collective.
- Returns:
- Return type:
See also
- getType()#
Return the type of the solver.
Not collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:192
- Return type:
- getUpdate()#
Return the callback to compute the update.
Not collective.
See also
- getVariableBounds()#
Return the lower and upper bounds vectors.
Not collective.
See also
- monitor(its=None, f=None, res=None, cnorm=None, step=None)#
Monitor the solver.
Collective.
This function should be called without arguments, unless users want to modify the values internally stored by the solver.
- Parameters:
its (int) – Current number of iterations or
Noneto use the value stored internally by the solver.f (float) – Current value of the objective function or
Noneto use the value stored internally by the solver.res (float) – Current value of the residual norm or
Noneto use the value stored internally by the solver.cnorm (float) – Current value of the constrains norm or
Noneto use the value stored internally by the solver.step (float) – Current value of the step or
Noneto use the value stored internally by the solver.
- Return type:
See also
- setALMMSubsolver(subsolver)#
Set the subsolver inside the ALMM solver.
Logically collective.
See also
- setALMMType(tao_almm_type)#
Set the ALMM type of the solver.
Logically collective.
See also
- setAppCtx(appctx)#
Set the application context.
- setBNCGType(cg_type)#
Set the type of the BNCG solver.
Collective.
See also
- setBRGNDictionaryMatrix(D)#
Set the dictionary matrix.
Collective.
See also
- setBRGNRegularizerHessian(hessian, H=None, args=None, kargs=None)#
Set the callback to compute the regularizer Hessian.
Logically collective.
See also
- setBRGNRegularizerObjectiveGradient(objgrad, args=None, kargs=None)#
Set the callback to compute the regularizer objective and gradient.
Logically collective.
- setBRGNRegularizerWeight(weight)#
Set the regularizer weight.
Collective.
- setBRGNSmoothL1Epsilon(epsilon)#
Set the smooth L1 epsilon.
Collective.
See also
- setConstraintTolerances(catol=None, crtol=None)#
Set the constraints tolerance parameters used in the solver convergence tests.
Collective.
- Parameters:
- Return type:
- setConstraints(constraints, C=None, args=None, kargs=None)#
Set the callback to compute constraints.
Logically collective.
- Parameters:
- Return type:
See also
- setConvergedReason(reason)#
Set the termination flag.
Collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:1305
- Parameters:
reason (ConvergedReason)
- Return type:
- setConvergenceTest(converged, args=None, kargs=None)#
Set the callback used to test for solver convergence.
Logically collective.
- Parameters:
- Return type:
See also
- setEqualityConstraints(equality_constraints, c, args=None, kargs=None)#
Set equality constraints callback.
Logically collective.
See also
- setFromOptions()#
Configure the solver from the options database.
Collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:248
- Return type:
- setGradient(gradient, g=None, args=None, kargs=None)#
Set the gradient evaluation callback.
Logically collective.
- Parameters:
- Return type:
See also
setObjective,setObjectiveGradient,setHessian,TaoSetGradient
- setGradientNorm(mat)#
Set the matrix used to compute inner products.
Collective.
See also
- setHessian(hessian, H=None, P=None, args=None, kargs=None)#
Set the callback to compute the Hessian matrix.
Logically collective.
- Parameters:
- Return type:
See also
getHessian,setObjective,setObjectiveGradient,setGradient,TaoSetHessian
- setInequalityConstraints(inequality_constraints, c, args=None, kargs=None)#
Set inequality constraints callback.
Logically collective.
See also
- setInitialTrustRegionRadius(radius)#
Set the initial trust region radius.
Collective.
See also
- setIterationNumber(its)#
Set the current iteration number.
Collective.
See also
- setJacobian(jacobian, J=None, P=None, args=None, kargs=None)#
Set the callback to compute the Jacobian.
Logically collective.
- Parameters:
- Return type:
See also
- setJacobianDesign(jacobian_design, J=None, args=None, kargs=None)#
Set Jacobian design callback.
Logically collective.
See also
- setJacobianEquality(jacobian_equality, J=None, P=None, args=None, kargs=None)#
Set Jacobian equality constraints callback.
Logically collective.
See also
- setJacobianInequality(jacobian_inequality, J=None, P=None, args=None, kargs=None)#
Set Jacobian inequality constraints callback.
Logically collective.
See also
- setJacobianResidual(jacobian, J=None, P=None, args=None, kargs=None)#
Set the callback to compute the least-squares residual Jacobian.
Logically collective.
- Parameters:
- Return type:
See also
- setJacobianState(jacobian_state, J=None, P=None, I=None, args=None, kargs=None)#
Set Jacobian state callback.
Logically collective.
See also
- setLMVMH0(mat)#
Set the initial Hessian for the quasi-Newton approximation.
Collective.
See also
- setLMVMMat(M)#
Set the LMVM matrix.
Logically collective.
See also
- setMaximumFunctionEvaluations(mit)#
Set the maximum number of objective evaluations within the solver.
Collective.
- setMaximumIterations(mit)#
Set the maximum number of solver iterations.
Collective.
See also
- setMonitor(monitor, args=None, kargs=None)#
Set the callback used to monitor solver convergence.
Logically collective.
- Parameters:
- Return type:
See also
- setObjective(objective, args=None, kargs=None)#
Set the objective function evaluation callback.
Logically collective.
- Parameters:
- Return type:
See also
- setObjectiveGradient(objgrad, g=None, args=None, kargs=None)#
Set the objective function and gradient evaluation callback.
Logically collective.
- Parameters:
- Return type:
- setOptionsPrefix(prefix)#
Set the prefix used for searching for options in the database.
Logically collective.
- setPythonContext(context)#
Set the instance of the class implementing the required Python methods.
Not collective.
- setPythonType(py_type)#
Set the fully qualified Python name of the class to be used.
Collective.
- setResidual(residual, R, args=None, kargs=None)#
Set the residual evaluation callback for least-squares applications.
Logically collective.
- Parameters:
- Return type:
See also
- setSolution(x)#
Set the vector used to store the solution.
Collective.
See also
- setStateDesignIS(state=None, design=None)#
Set the index sets indicating state and design variables.
Collective.
See also
- setTolerances(gatol=None, grtol=None, gttol=None)#
Set the tolerance parameters used in the solver convergence tests.
Collective.
- Parameters:
gatol (float) – The absolute norm of the gradient, or
DETERMINEto use the value when the object’s type was set. Defaults toCURRENT.grtol (float) – The relative norm of the gradient with respect to the initial norm of the objective, or
DETERMINEto use the value when the object’s type was set. Defaults toCURRENT.gttol (float) – The relative norm of the gradient with respect to the initial norm of the gradient, or
DETERMINEto use the value when the object’s type was set. Defaults toCURRENT.
- Return type:
See also
- setType(tao_type)#
Set the type of the solver.
Logically collective.
See also
- setUp()#
Set up the internal data structures for using the solver.
Collective.
See also
Source code at petsc4py/PETSc/TAO.pyx:260
- Return type:
- setUpdate(update, args=None, kargs=None)#
Set the callback to compute update at each optimization step.
Logically collective.
- Parameters:
- Return type:
See also
- setVariableBounds(varbounds, args=None, kargs=None)#
Set the lower and upper bounds for the optimization problem.
Logically collective.
- Parameters:
- Return type:
- solve(x=None)#
Solve the optimization problem.
Collective.
- Parameters:
x (Vec | None) – The starting vector or
Noneto use the vector stored internally.- Return type:
See also
- view(viewer=None)#
View the solver.
Collective.
See also
Attributes Documentation
- appctx#
Application context.
- cnorm#
Constraints norm.
- converged#
Boolean indicating if the solver has converged.
- ctol#
Broken.
- diverged#
Boolean indicating if the solver has failed.
- ftol#
Broken.
- function#
Objective value.
- gnorm#
Gradient norm.
- gradient#
Gradient vector.
- gtol#
Broken.
- iterating#
Boolean indicating if the solver has not converged yet.
- its#
Number of iterations.
- ksp#
Linear solver.
- objective#
Objective value.
- reason#
Converged reason.
- solution#
Solution vector.