torch::autograd::SavedVariable Class Reference

A snapshot of a variable at a certain version. More...

#include <saved_variable.h>

Public Member Functions
	SavedVariable (const Variable &variable, bool is_output)
	SavedVariable (SavedVariable &&)=default
SavedVariable &	operator= (SavedVariable &&)=default
Variable	unpack (std::shared_ptr< Function > saved_for=nullptr) const
	Reconstructs the saved variable. More...
void	reset_data ()
void	reset_grad_function ()

Detailed Description

A snapshot of a variable at a certain version.

A SavedVariable stores enough information to reconstruct a variable from a certain point in time.

Definition at line 21 of file saved_variable.h.

Member Function Documentation

Variable torch::autograd::SavedVariable::unpack

(

std::shared_ptr< Function >

saved_for = nullptr

)

const

Reconstructs the saved variable.

Pass saved_for as the gradient function if constructing the SavedVariable with it would have caused a circular reference.

Definition at line 34 of file saved_variable.cpp.

---

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

• torch/csrc/autograd/saved_variable.h
• torch/csrc/autograd/saved_variable.cpp