torch::jit::script::Module Struct Reference

Public Member Functions
	TH_DISALLOW_COPY_AND_ASSIGN (Module)
void	set_optimized (bool o)
bool	is_optimized () const
IValue	forward (std::vector< IValue > inputs)
void	register_buffer (const std::string &name, autograd::Variable v)
void	register_parameter (const std::string &name, autograd::Variable v, bool is_buffer)
void	register_attribute (const std::string &name, const TypePtr type, IValue ivalue)
void	register_module (const std::string &name, std::shared_ptr< Module > module)
Method &	create_method (const std::string &name, std::shared_ptr< Graph > graph, std::vector< IValue * > member_inputs)
Method &	create_method (const std::string &name, std::function< void(Method &)> creator)
IValue *	parameter_slot (const std::string &name) const
void	set_parameter (const std::string &name, at::Tensor v)
autograd::Variable	get_parameter (const std::string &name) const
autograd::Variable	get_buffer (const std::string &name) const
Method &	get_method (const std::string &name) const
std::shared_ptr< Module >	get_module (const std::string &name) const
const torch::OrderedDict< std::string, NamedModule > &	get_modules () const
const torch::OrderedDict< std::string, NamedIValue > &	get_parameters () const
const torch::OrderedDict< std::string, NamedIValue > &	get_attributes () const
const torch::OrderedDict< std::string, std::unique_ptr< Method > > &	get_methods () const
NamedIValue *	find_parameter (const std::string &name)
NamedIValue *	find_attribute (const std::string &name)
NamedIValue *	find_buffer (const std::string &name)
NamedModule *	find_module (const std::string &name)
Method *	find_method (const std::string &name)
void	apply (std::function< void(Module &)> fn)
void	train (bool on=true)
	Enables "training" mode.
void	eval ()
	Calls train(false) to enable "eval" mode. More...
bool	is_training ()
	True if the module is in training mode.
TORCH_API void	to (at::Device device, at::ScalarType dtype, bool non_blocking=false)
	Recursively casts all parameters to the given dtype and device. More...
TORCH_API void	to (at::ScalarType dtype, bool non_blocking=false)
	Recursively casts all parameters to the given dtype. More...
TORCH_API void	to (at::Device device, bool non_blocking=false)
	Recursively moves all parameters to the given device. More...
template<typename... Types>
IValue	run_method (const std::string &method_name, Types &&...args)
	Run a method from this module. More...
void	save (std::ostream &out, const ExtraFilesMap &extra_files=ExtraFilesMap())
void	save (const std::string &filename, const ExtraFilesMap &extra_files=ExtraFilesMap())
void	copy_into (ModuleLookup module_lookup, std::unordered_map< IValue *, IValue * > &parameter_remap, std::vector< std::string > names={}) const

Detailed Description

Definition at line 375 of file module.h.

Member Function Documentation

void torch::jit::script::Module::eval ( )

inline

Calls train(false) to enable "eval" mode.

Do not override this method, override train() instead.

Definition at line 538 of file module.h.

template<typename... Types>

IValue torch::jit::script::Module::run_method ( const std::string &  method_name, Types &&...  args  )

inline

Run a method from this module.

For example:

IValue output = module->run("relu_script", a, b);

To get a compile a module from a source string, see torch::jit::compile

Parameters

method_name	The name of the method to run
args	Arguments to be passed to the method

Returns

An IValue containing the return value (or values if it is a tuple) from the method

Definition at line 591 of file module.h.

void torch::jit::script::Module::to	(	at::Device	device,
		at::ScalarType	dtype,
		bool	non_blocking = false
	)

Recursively casts all parameters to the given dtype and device.

If non_blocking is true and the source is in pinned memory and destination is on the GPU or vice versa, the copy is performed asynchronously with respect to the host. Otherwise, the argument has no effect.

Definition at line 98 of file module.cpp.

void torch::jit::script::Module::to	(	at::ScalarType	dtype,
		bool	non_blocking = false
	)

Recursively casts all parameters to the given dtype.

If non_blocking is true and the source is in pinned memory and destination is on the GPU or vice versa, the copy is performed asynchronously with respect to the host. Otherwise, the argument has no effect.

Definition at line 102 of file module.cpp.

void torch::jit::script::Module::to	(	at::Device	device,
		bool	non_blocking = false
	)

Recursively moves all parameters to the given device.

If non_blocking is true and the source is in pinned memory and destination is on the GPU or vice versa, the copy is performed asynchronously with respect to the host. Otherwise, the argument has no effect.

Definition at line 106 of file module.cpp.

---

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

• torch/csrc/jit/script/module.h
• torch/csrc/jit/script/module.cpp