Struct Tensor 

pub struct Tensor {
    pub data: Vec<f32>,
    pub shape: Vec<usize>,
}

N-dimensional tensor with f32 data and a shape vector. Layout: row-major (C order), last dimension varies fastest.

Fields

data: Vec<f32>

shape: Vec<usize>

Implementations

impl Tensor

pub fn new(data: Vec<f32>, shape: Vec<usize>) -> Tensor

Create a tensor from owned data and a shape vector. Panics (debug-only) if data.len() != shape.product().

pub fn zeros(shape: &[usize]) -> Tensor

Create a zero-filled tensor with the given shape.

pub fn scalar(val: f32) -> Tensor

Create a scalar tensor (shape [1]) containing a single value.

pub fn numel(&self) -> usize

Total number of elements in this tensor.

pub fn ndim(&self) -> usize

Number of dimensions (rank) of this tensor.

pub fn reshape(&self, new_shape: &[usize]) -> Tensor

Return a new tensor with the same data but a different shape. Panics if the element count changes.

pub fn broadcast_shape(a: &[usize], b: &[usize]) -> Result<Vec<usize>, String>

Compute the broadcast shape of two tensors (NumPy rules). Returns Err if shapes are incompatible.

pub fn get(&self, idx: usize) -> f32

Retrieve a single element by flat index (bounds checked in debug).

impl Tensor

pub fn view(&self) -> TensorView<'_>

Create a contiguous view of this tensor.

pub fn transpose_view(&self, perm: &[usize]) -> TensorView<'_>

Create a transposed view without copying data.

pub fn slice_view( &self, axis: usize, start: usize, end: usize, ) -> TensorView<'_>

Create a sliced view without copying data.

impl Tensor

pub fn broadcast_iter<'a>( &'a self, other: &'a Tensor, ) -> Option<BroadcastIter<'a>>

Create a broadcast iterator pairing this tensor with another.

Trait Implementations

impl Clone for Tensor

fn clone(&self) -> Tensor

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for Tensor

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

Formats the value using the given formatter.