Struct DecoderStack 

pub struct DecoderStack {
    pub config: DecoderStackConfig,
    pub layers: Vec<DecoderLayer>,
    pub position_encoding_sin: Option<SinusoidalPositionEncoding>,
    pub position_encoding_learned: Option<LearnedPositionEncoding>,
    pub final_norm: Option<LayerNorm>,
}

Transformer decoder stack

Fields

config: DecoderStackConfig

Configuration

layers: Vec<DecoderLayer>

Decoder layers

position_encoding_sin: Option<SinusoidalPositionEncoding>

Position encoding (if sinusoidal)

position_encoding_learned: Option<LearnedPositionEncoding>

Position encoding (if learned)

final_norm: Option<LayerNorm>

Final layer normalization

Implementations

impl DecoderStack

pub fn new(config: DecoderStackConfig) -> Result<Self>

Create a new decoder stack

pub fn build_decoder_stack_graph( &self, graph: &mut EinsumGraph, ) -> Result<Vec<usize>>

Build einsum graph for decoder stack

Input tensors:

• 0: x (target input) [batch, tgt_len, d_model]
• 1: encoder_output [batch, src_len, d_model]
• 2-N: all parameters

Output tensors:

• output: [batch, tgt_len, d_model]

pub fn num_layers(&self) -> usize

Get number of layers

Trait Implementations

impl Clone for DecoderStack

fn clone(&self) -> DecoderStack

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for DecoderStack

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

Formats the value using the given formatter.