Struct TransformerMetaLearning 

pub struct TransformerMetaLearning<T: Float + Debug + ScalarOperand + Send + Sync + 'static> { /* private fields */ }

Transformer meta-learning implementation

Implementations

impl<T: Float + Debug + ScalarOperand + Send + Sync + 'static> TransformerMetaLearning<T>

pub fn new(config: &TransformerBasedOptimizerConfig<T>) -> Result<Self>

Create new transformer meta-learning component

pub fn meta_step( &mut self, tasks: &[TaskBatch<T>], support_data: &[Array2<T>], query_data: &[Array2<T>], ) -> Result<MetaLearningResult<T>>

Perform meta-learning step

pub fn generate_update( &mut self, transformer_output: &Array2<T>, current_parameters: &Array1<T>, ) -> Result<Array1<T>>

Generate optimization update using meta-learned parameters

pub fn update_from_loss(&mut self, loss: T) -> Result<()>

Update meta-learning state from loss

pub fn parameter_count(&self) -> usize

Total learnable parameters held by the adaptation networks.

pub fn meta_optimizer_mut(&mut self) -> &mut MetaOptimizer<T>

Mutable access to the outer-loop optimizer (e.g. to enable momentum).

pub fn set_strategy(&mut self, strategy: MetaLearningStrategy)

Set meta-learning strategy

pub fn get_strategy(&self) -> MetaLearningStrategy

Get current strategy