mxnet::TensorInspector Class Reference

This class provides a unified interface to inspect the value of all data types including Tensor, TBlob, and NDArray. If the tensor resides on GPU, then it will be copied from GPU memory back to CPU memory to be operated on. Internally, all data types are stored as a TBlob object tb_. More...

#include <tensor_inspector.h>

Collaboration diagram for mxnet::TensorInspector:

Public Member Functions
template<typename Device , int dimension, typename DType >
	TensorInspector (const mshadow::Tensor< Device, dimension, DType > &ts, const RunContext &ctx)
	construct from Tensor object More...
	TensorInspector (const TBlob &tb, const RunContext &ctx)
	construct from TBlob object More...
	TensorInspector (const NDArray &arr, const RunContext &ctx)
	construct from NDArray object. Currently this only works with kDefaultStorage More...
void	print_string ()
	print the tensor to std::cout More...
std::string	to_string ()
	return a string which contains the values and other info of the tensor More...
void	interactive_print (std::string tag="")
	interactively print the tensor value More...
template<typename ValueChecker >
std::vector< std::vector< index_t > >	check_value (const ValueChecker &checker, bool interactive=false, std::string tag="")
	check/validate the values within the tensor, return the coordinates where the value checker evaluates to true More...
std::vector< std::vector< index_t > >	check_value (CheckerType ct, bool interactive=false, std::string tag="")
	check/validate the values within the tensor, return the coordinates where the lambda evaluates to true More...
void	dump_to_file (std::string tag)
	dump the value of the tensor to a file with name "tag_[visit count].npy" in npy format More...

Detailed Description

This class provides a unified interface to inspect the value of all data types including Tensor, TBlob, and NDArray. If the tensor resides on GPU, then it will be copied from GPU memory back to CPU memory to be operated on. Internally, all data types are stored as a TBlob object tb_.

---

|__ __| |_ _| | | | | ___ _ __ ___ ___ _ __| | _ __ ___ _ __ ___ ___| |_ ___ _ __ | |/ _ \ '_ \/ __|/ _ | '__| | | '_ \/ __| '_ \ / _ \/ __| __/ _ | '__| | | __/ | | __ \ (_) | | _| |_| | | __ \ |_) | __/ (__| || (_) | | |_|___|_| |_|___/___/|_||_____|_| |_|___/ .__/ ___|___|_____/|_| | | |_|

Constructor & Destructor Documentation

template<typename Device , int dimension, typename DType >

mxnet::TensorInspector::TensorInspector ( const mshadow::Tensor< Device, dimension, DType > &  ts, const RunContext &  ctx  )

inline

construct from Tensor object

Template Parameters

Device	the device the tensor resides in
dimension	the dimension of the tensor
DType	the data type

Parameters

ts	the source tensor object
ctx	the run context of the tensor

mxnet::TensorInspector::TensorInspector ( const TBlob &  tb, const RunContext &  ctx  )

inline

construct from TBlob object

Parameters

tb	the source tblob object
ctx	the run context of the tensor

mxnet::TensorInspector::TensorInspector ( const NDArray &  arr, const RunContext &  ctx  )

inline

construct from NDArray object. Currently this only works with kDefaultStorage

Parameters

arr	the source ndarray object
ctx	the run context of the tensor

Member Function Documentation

template<typename ValueChecker >

std::vector<std::vector<index_t> > mxnet::TensorInspector::check_value ( const ValueChecker &  checker, bool  interactive = false, std::string  tag = ""  )

inline

check/validate the values within the tensor, return the coordinates where the value checker evaluates to true

Template Parameters

ValueChecker

the type of the lambda

Parameters

checker	the lambda function to check each value of within the tensor
interactive	wherether to allow the user to interactively check the coordinates
tag	the name given to this call

std::vector<std::vector<index_t> > mxnet::TensorInspector::check_value ( CheckerType  ct, bool  interactive = false, std::string  tag = ""  )

inline

check/validate the values within the tensor, return the coordinates where the lambda evaluates to true

Parameters

ct	the type of the checker
interactive	wherether to allow the user to interactively check the coordinates
tag	the name given to this call

void mxnet::TensorInspector::dump_to_file ( std::string  tag )

inline

dump the value of the tensor to a file with name "tag_[visit count].npy" in npy format

Parameters

tag	the name given to this call

void mxnet::TensorInspector::interactive_print ( std::string  tag = "" )

inline

interactively print the tensor value

Parameters

tag	the name given to this call

void mxnet::TensorInspector::print_string ( )

inline

print the tensor to std::cout

std::string mxnet::TensorInspector::to_string ( )

inline

return a string which contains the values and other info of the tensor

---

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

• /work/mxnet/src/common/tensor_inspector.h