Struct WeightTensor 

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

A named flat f32 weight tensor.

Weights are stored in row-major order. For a 2-D matrix W[m, n], the element at row i column j is at data[i * n + j].

Fields

data: Vec<f32>

Raw data in row-major order.

shape: Vec<usize>

Shape of the tensor (product equals data.len()).

Implementations

impl WeightTensor

pub fn from_data(data: Vec<f32>, shape: Vec<usize>) -> LmResult<Self>

Construct from existing data, validating that data.len() equals the product of shape.

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

Tensor filled with zeros.

pub fn ones(shape: &[usize]) -> Self

Tensor filled with ones.

pub fn eye(rows: usize, cols: usize) -> Self

Identity-like weight: ones on the “diagonal” of a 2-D matrix.

For a non-square matrix the identity is placed in the top-left corner.

pub fn n_elements(&self) -> usize

Total number of elements.

pub fn ndim(&self) -> usize

Number of tensor dimensions.

pub fn as_slice(&self) -> &[f32]

Borrow the underlying data.

pub fn as_mut_slice(&mut self) -> &mut [f32]

Mutable borrow of the underlying data.

pub fn row_slice(&self, row: usize) -> LmResult<&[f32]>

For a 2-D weight matrix [rows × cols], return the row-th row slice.

pub fn validate_shape(&self, expected: &[usize]) -> LmResult<()>

Validate that this tensor has the given shape.

Trait Implementations

impl Clone for WeightTensor

fn clone(&self) -> WeightTensor

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for WeightTensor

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

Formats the value using the given formatter.