Trait DynamicModel 

pub trait DynamicModel {
    // Required methods
    fn forward_dynamic(
        &self,
        input: &ArrayView<'_, f64, IxDyn>,
    ) -> TrainResult<Array<f64, IxDyn>>;
    fn backward_dynamic(
        &self,
        input: &ArrayView<'_, f64, IxDyn>,
        grad_output: &ArrayView<'_, f64, IxDyn>,
    ) -> TrainResult<HashMap<String, Array<f64, IxDyn>>>;
}

Trait for models with dynamic computation graphs.

This extends the model interface to support variable-sized inputs and dynamic graph construction (e.g., for RNNs, variable-length sequences).

Required Methods

fn forward_dynamic( &self, input: &ArrayView<'_, f64, IxDyn>, ) -> TrainResult<Array<f64, IxDyn>>

Forward pass with dynamic input dimensions.

fn backward_dynamic( &self, input: &ArrayView<'_, f64, IxDyn>, grad_output: &ArrayView<'_, f64, IxDyn>, ) -> TrainResult<HashMap<String, Array<f64, IxDyn>>>

Backward pass with dynamic input dimensions.

Implementors