Struct LSTMCell

pub struct LSTMCell {
    pub w_ih: Array2<f64>,
    pub w_hh: Array2<f64>,
    pub b_ih: Array2<f64>,
    pub b_hh: Array2<f64>,
    pub hidden_size: usize,
    pub input_dropout: Option<Dropout>,
    pub recurrent_dropout: Option<Dropout>,
    pub output_dropout: Option<Dropout>,
    pub zoneout: Option<Zoneout>,
    pub is_training: bool,
}

LSTM cell with trainable parameters and dropout support

Fields

w_ih: Array2<f64>

w_hh: Array2<f64>

b_ih: Array2<f64>

b_hh: Array2<f64>

hidden_size: usize

input_dropout: Option<Dropout>

recurrent_dropout: Option<Dropout>

output_dropout: Option<Dropout>

zoneout: Option<Zoneout>

is_training: bool

Implementations

impl LSTMCell

pub fn new(input_size: usize, hidden_size: usize) -> Self

Creates new LSTM cell with Xavier-uniform weight initialization

pub fn with_input_dropout(self, dropout_rate: f64, variational: bool) -> Self

pub fn with_recurrent_dropout( self, dropout_rate: f64, variational: bool, ) -> Self

pub fn with_output_dropout(self, dropout_rate: f64) -> Self

pub fn with_zoneout( self, cell_zoneout_rate: f64, hidden_zoneout_rate: f64, ) -> Self

pub fn train(&mut self)

pub fn eval(&mut self)

pub fn forward( &mut self, input: &Array2<f64>, hx: &Array2<f64>, cx: &Array2<f64>, ) -> (Array2<f64>, Array2<f64>)

pub fn forward_with_cache( &mut self, input: &Array2<f64>, hx: &Array2<f64>, cx: &Array2<f64>, ) -> (Array2<f64>, Array2<f64>, LSTMCellCache)

pub fn forward_batch( &mut self, input: &Array2<f64>, hx: &Array2<f64>, cx: &Array2<f64>, ) -> (Array2<f64>, Array2<f64>)

Batch forward pass for multiple sequences simultaneously

Arguments

• input - Input tensor of shape (input_size, batch_size)
• hx - Hidden state tensor of shape (hidden_size, batch_size)
• cx - Cell state tensor of shape (hidden_size, batch_size)

Returns

• Tuple of (new_hidden_state, new_cell_state) with same batch dimensions

pub fn forward_batch_with_cache( &mut self, input: &Array2<f64>, hx: &Array2<f64>, cx: &Array2<f64>, ) -> (Array2<f64>, Array2<f64>, LSTMCellBatchCache)

Batch forward pass with caching for training

Similar to forward_batch but caches intermediate values needed for backpropagation

pub fn backward( &self, dhy: &Array2<f64>, dcy: &Array2<f64>, cache: &LSTMCellCache, ) -> (LSTMCellGradients, Array2<f64>, Array2<f64>, Array2<f64>)

Backward pass implementing LSTM gradient computation with dropout

Returns (parameter_gradients, input_gradient, hidden_gradient, cell_gradient)

pub fn backward_batch( &self, dhy: &Array2<f64>, dcy: &Array2<f64>, cache: &LSTMCellBatchCache, ) -> (LSTMCellGradients, Array2<f64>, Array2<f64>, Array2<f64>)

Batch backward pass for training with multiple sequences

Computes gradients for an entire batch simultaneously

pub fn zero_gradients(&self) -> LSTMCellGradients

Initialize zero gradients for accumulation

pub fn update_parameters<O: Optimizer>( &mut self, gradients: &LSTMCellGradients, optimizer: &mut O, prefix: &str, )

Apply gradients using the provided optimizer

Trait Implementations

impl Clone for LSTMCell

fn clone(&self) -> LSTMCell

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl From<&LSTMCell> for SerializableLSTMCell

fn from(cell: &LSTMCell) -> Self

Converts to this type from the input type.

impl Into<LSTMCell> for SerializableLSTMCell

fn into(self) -> LSTMCell

Converts this type into the (usually inferred) input type.