TensorBase

Function Usage

Functions as the TensorBase constructor, which is used to create TensorBase objects. Different constructors can be selected based on the passed parameters.

If an exception indicating that the constructor fails to be executed due to memory exhaustion occurs, do not call subsequent member functions.

Prototype

Prototype 1:

TensorBase::TensorBase();

Prototype 2:

TensorBase::TensorBase(const MemoryData &memoryData, const bool &isBorrowed, const std::vector<uint32_t> &shape, const TensorDataType &type);

Prototype 3:

TensorBase::TensorBase(const std::vector<uint32_t> &shape, const TensorDataType &type, const MemoryData::MemoryType &bufferType, const int32_t &deviceId);

Prototype 4:

TensorBase::TensorBase(const std::vector<uint32_t> &shape, const TensorDataType &type, const int32_t &deviceId);

Prototype 5:

TensorBase::TensorBase(const std::vector<uint32_t> &shape, const TensorDataType &type);

Prototype 6:

TensorBase::TensorBase(const std::vector<uint32_t> &shape);

Prototype 7:

TensorBase::TensorBase(const TensorBase& tensor) = default;

Parameters

The prototypes 3, 4, 5, and 6 only preset the shapes, but do not allocate the memory spaces. You need to call TensorBaseMalloc to allocate the corresponding memory space.
To use the prototype 2, allocate the external memory space, and then directly use the allocated memory to construct the tensor object. Ensure that the space size is the same as the shape of the tensor object and determine whether the externally allocated space is destroyed by the tensor object or by the user.

Parameter	Input/Output	Description
memoryData	Input	Memory management structure, a parameter used to construct the TensorBase object. For details, see MemoryData.
isBorrowed	Input	Whether the input memory data needs to be proactively destroyed by the tensor. If the value is true, the data does not need to be proactively destroyed, but is manually destroyed by the user. If the value is false, the data does not need to be manually destroyed by the user, but is automatically destroyed during tensor destruction.
shape	Input	Tensor shape, a parameter used to construct the TensorBase object.
type	Input	Tensor data type, a parameter used to construct the TensorBase object. For details, see the enumerations of TensorDataType.
bufferType	Input	Memory type of the tensor data, a parameter used to construct the TensorBase object.
deviceId	Input	Device ID of the int type, a parameter used to construct the TensorBase object.
tensor	Input	TensorBase object, a parameter used to construct the TensorBase object.

Parent topic: TensorBase