Struct TransformerBlock 

pub struct TransformerBlock {
    pub mha: MultiHeadAttention,
    pub ffn: TransformerFfn,
    pub ln1: LayerNorm,
    pub ln2: LayerNorm,
    pub d_model: usize,
}

A single transformer encoder block: x → MHA(LayerNorm(x)) + x → FFN(LayerNorm(·)) + ·

Fields

mha: MultiHeadAttention

Multi-head self-attention module.

ffn: TransformerFfn

Feed-forward network.

ln1: LayerNorm

Layer norm before MHA.

ln2: LayerNorm

Layer norm before FFN.

d_model: usize

Model dimensionality.

Implementations

impl TransformerBlock

pub fn new(d_model: usize, n_heads: usize, d_ff: usize) -> Self

Create a new transformer block with zero weights.

pub fn forward(&self, x: &[f64], seq_len: usize) -> Vec<f64>

Forward pass with pre-norm style residual connections.

x is flat row-major [seq_len × d_model].

Trait Implementations

impl Clone for TransformerBlock

fn clone(&self) -> TransformerBlock

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TransformerBlock

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

Formats the value using the given formatter.