torch::optim::LossClosureOptimizer Class Reference

Optimizer that requires the loss function to be supplied to the step() function, as it may evaluate the loss function multiple times per step. More...

#include <optimizer.h>

Inheritance diagram for torch::optim::LossClosureOptimizer:

Public Types
using	LossClosure = std::function< Tensor()>
	A loss function closure, which is expected to return the loss value.

Public Member Functions
virtual Tensor	step (LossClosure closure)=0
Public Member Functions inherited from torch::optim::detail::OptimizerBase
	OptimizerBase (std::vector< Tensor > parameters)
	Constructs the Optimizer from a vector of parameters.
void	add_parameters (const std::vector< Tensor > &parameters)
	Adds the given vector of parameters to the optimizer's parameter list.
virtual void	zero_grad ()
	Zeros out the gradients of all parameters.
const std::vector< Tensor > &	parameters () const noexcept
	Provides a const reference to the parameters this optimizer holds.
std::vector< Tensor > &	parameters () noexcept
	Provides a reference to the parameters this optimizer holds.
size_t	size () const noexcept
	Returns the number of parameters referenced by the optimizer.
virtual void	save (serialize::OutputArchive &archive) const
	Serializes the optimizer state into the given archive.
virtual void	load (serialize::InputArchive &archive)
	Deserializes the optimizer state from the given archive.

Additional Inherited Members
Protected Member Functions inherited from torch::optim::detail::OptimizerBase
template<typename T >
T &	buffer_at (std::vector< T > &buffers, size_t index)
	Accesses a buffer at the given index. More...
Tensor &	buffer_at (std::vector< Tensor > &buffers, size_t index)
	Accesses a buffer at the given index, converts it to the type of the parameter at the corresponding index (a no-op if they match). More...
Protected Attributes inherited from torch::optim::detail::OptimizerBase
std::vector< Tensor >	parameters_
	The parameters this optimizer optimizes.

Detailed Description

Optimizer that requires the loss function to be supplied to the step() function, as it may evaluate the loss function multiple times per step.

Examples of such algorithms are conjugate gradient and LBFGS. The step() function also returns the loss value.

Definition at line 110 of file optimizer.h.

---

The documentation for this class was generated from the following file:

• torch/csrc/api/include/torch/optim/optimizer.h