Tensor

class tensorrt.plugin.Tensor

Representation of a tensor that carries data

Tensor objects are strictly descriptors of a tensor with an underlying data buffer. tensorrt.plugin does not provide any APIs that perform standard data-altering operations on Tensors.

Supports __cuda_array_interface__ for interoperability with other frameworks.

aliased(shape: Shape | Tuple[int] | PluginTensorDesc | None = None) Tensor

Return a Tensor which has the same data_ptr as this but has the provided shape.

Parameters:

shape (Union[Shape, Tuple[int], trt.PluginTensorDesc], optional) – Required shape of the new tensor (must have the same volume). Defaults to same shape.

Raises:

ValueError – If shape is not a supported type or if it does not have the same volume

property data_ptr: int

Pointer to the data buffer of this tensor

property dtype: DataType

Data type of the tensor.

property format: TensorFormat

The format of the tensor.

get_aliased() Tensor
Returns:

Returns Tensor of the tensor which this tensor is aliased to. Returns None is this tensor is not aliased to any other tensor.

property is_aliased

True if this tensor is aliased to another tensor, False otherwise.

property is_data_dependent

True if this tensor contains at least one data-dependent dimension, False otherwise.

property ndim: int

Number of dimensions

numel() int

Returns the number of elements of the tensor

Raises:

ValueError – If the tensor has a data-dependent dimension. Examine is_data_dependent to determine whether the tensor is data-dependent.

Returns:

Number of elements of the tensor

Return type:

int

property scale: float

Scale for INT8 data type.

property shape: Shape

The (concrete) shape of the tensor.

property strides: Tuple[int]

Strides of this tensor.