lmp::tensor::TensorImpl Class Reference

Main implementation class for Tensor object. More...

#include <tensor_impl.hpp>

Public Member Functions
template<typename T >
	TensorImpl (const std::vector< T > &data, const std::vector< size_t > &shape, DeviceType device, DataType dtype)
	Construct a TensorImpl from a vector of data.
	TensorImpl (Storage storage, const std::vector< size_t > &shape, DataType dtype)
void *	data () const noexcept
DataType	type () const noexcept
DeviceType	device () const noexcept
const std::vector< size_t > &	shape () const noexcept
const std::vector< detail::stride_t > &	strides () const noexcept
size_t	numel () const noexcept
TensorImpl	reshape (std::vector< size_t > new_shape)
TensorImpl	squeeze (size_t dim)
TensorImpl	expand_dims (size_t dim)
Scalar	index (const std::vector< size_t > &idx)
void	copy (const TensorImpl &other) const
void	fill (Scalar item) const
void	print (std::ostream &os) const

Friends
class	Tensor

Detailed Description

Main implementation class for Tensor object.

TensorImpl contains a few core members: type_, shape_, and data_ Note that similar to Pytorch, Tensor/TensorImpl is not responsible for the low-level data storage – note that TensorImpl has no member called device_. That is managed by Storage.

See also

Tensor, Storage

Constructor & Destructor Documentation

TensorImpl()

template<typename T >

lmp::tensor::TensorImpl::TensorImpl ( const std::vector< T > &  data, const std::vector< size_t > &  shape, DeviceType  device, DataType  dtype  )

inlineexplicit

Construct a TensorImpl from a vector of data.

Template Parameters

T	The element type of the input data vector

Parameters

data	Flat vector containing the tensor data in row-major order
shape	Dimensions of the tensor, e.g. {28, 28} for a 2D tensor
device	Target device where the tensor will be stored (CPU/GPU)
dtype	Data type for the tensor elements (may differ from T)

Exceptions

std::runtime_error

if data.size() != product of shape dimensions

This constructor allocates storage on the specified device and copies the input data.

Note

Note that the input data's type T does NOT have to be the same as dtype. i.e. inputting dtype = DataType::Float64, but data = std::vector<int>{...} is valid

---

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

• csrc/include/lamppp/tensor/tensor_impl.hpp
• csrc/src/tensor/tensor_impl.cpp