Class TensorBase¶

Defined in File tensor.h

Class Documentation¶

class TensorBase¶

Main implementation of a tensor, a multi-dimensional matrix containing elements of a single data type.

TensorBase contains the data, data type, pointer to memory region, shape, backend info and other attributes.

Public Types

typedef IPtr<TensorBase> PtrType¶

Public Functions

virtual ~TensorBase()¶

virtual void reset(MemoryPiece::PtrType memory)¶

virtual MemoryPiece::PtrType memory()¶

virtual Type type()¶

virtual Shape &shape()¶

virtual float *data()¶

template<typename T> T *data()¶

virtual size_t size()¶

template<typename T> T scalar()¶

virtual float scalar()¶

Ptr<Backend> getBackend()¶

DeviceId getDeviceId()¶

Tensor subtensor(size_t offset, size_t size)¶

template<typename T> T get(size_t i)¶

float get(size_t i)¶

template<typename T> void get(std::vector<T> &v)¶

void get(io::Item &item, const std::string &name)¶

template<typename T> void set(size_t i, T value)¶

template<typename T> void set(const T *begin, const T *end)¶

template<typename T> void set(const std::vector<T> &v)¶

void set(const char *begin, const char *end, Type type)¶

void set(const std::vector<char> &v, Type type)¶

void set(const io::Item &item)¶

template<typename T> void set(T value)¶

void setSparse(const std::vector<size_t> &k, const std::vector<float> &v)¶

template<typename T> void copyFrom(Tensor in)¶

void copyFrom(Tensor in)¶

template<typename T> void swap(Tensor swapee)¶

void swap(Tensor swapee)¶

template<typename T> std::string debug(int precision = 8, int dispCols = 5)¶

template std::string marian::TensorBase::debug< int64_t >(int precision = 8, int dispCols = 5)

size_t hash()¶

Public Static Functions

template<class ...Args> static PtrType New(Args&&... args)¶