Enum DynTensor 

pub enum DynTensor<D: Device> {
    Float(Tensor<D, Float>),
    Int(Tensor<D, Int>),
    Bool(Tensor<D, Bool>),
}

A tensor whose kind (Float/Int/Bool) is determined at runtime.

This is the type-erased counterpart to Tensor<D, K>. Use it when you need to store or pass tensors of different kinds together — most importantly for disk I/O (from_bytes / to_bytes).

Variants

Float(Tensor<D, Float>)

Int(Tensor<D, Int>)

Bool(Tensor<D, Bool>)

Implementations

impl<D: Device> DynTensor<D>

pub fn dtype(&self) -> DType

Runtime element type.

pub fn shape(&self) -> &Shape

Logical shape.

pub fn dims(&self) -> &[usize]

Dimensions as a slice.

pub fn device(&self) -> &D

Device that owns the storage.

impl<D: Device> DynTensor<D>

pub fn as_float(&self) -> Option<&Tensor<D, Float>>

Borrow as a Float tensor, or None.

pub fn as_int(&self) -> Option<&Tensor<D, Int>>

Borrow as an Int tensor, or None.

pub fn as_bool(&self) -> Option<&Tensor<D, Bool>>

Borrow as a Bool tensor, or None.

pub fn into_float(self) -> Result<Tensor<D, Float>>

Unwrap into a Float tensor, or error.

pub fn into_int(self) -> Result<Tensor<D, Int>>

Unwrap into an Int tensor, or error.

pub fn into_bool(self) -> Result<Tensor<D, Bool>>

Unwrap into a Bool tensor, or error.

impl<D: Device> DynTensor<D>

pub fn from_bytes( bytes: &[u8], dtype: DType, shape: impl Into<Shape>, device: &D, ) -> Result<Self>

Create a DynTensor from raw little-endian bytes and a runtime DType.

This is the primary entry point for deserialization (safetensors, NPY, …). The byte slice length must equal shape.element_count() * dtype.size_in_bytes().

pub fn to_bytes(&self) -> Result<Vec<u8>>

Serialize to raw little-endian bytes in logical order.

Trait Implementations

impl<D: Device> Clone for DynTensor<D>

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl<D: Device> Display for DynTensor<D>

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<D: Device> From<Tensor<D, Bool>> for DynTensor<D>

fn from(t: Tensor<D, Bool>) -> Self

Converts to this type from the input type.

impl<D: Device> From<Tensor<D, Int>> for DynTensor<D>

fn from(t: Tensor<D, Int>) -> Self

Converts to this type from the input type.

impl<D: Device> From<Tensor<D>> for DynTensor<D>

fn from(t: Tensor<D, Float>) -> Self

Converts to this type from the input type.