Struct FusedOptimizer 

pub struct FusedOptimizer { /* private fields */ }

Fused optimizer that combines multiple optimizers for efficiency

Implementations

impl FusedOptimizer

pub fn new( optimizers: Vec<Box<dyn Optimizer>>, config: FusionConfig, ) -> Result<Self>

Create a new fused optimizer

pub fn fused_step( &mut self, parameters: &mut HashMap<String, Tensor>, ) -> Result<()>

Perform fused optimization step

pub fn register_gradient( &self, param_name: &str, gradient: Tensor, ) -> Result<()>

Register gradient for parameter in the fused gradient registry

This method allows external automatic differentiation systems to register computed gradients with the fused optimizer for parameter updates.

pub fn clear_gradients(&self) -> Result<()>

Clear all registered gradients

This should be called after each optimization step to clear accumulated gradients.

pub fn get_available_gradient_names(&self) -> Result<Vec<String>>

Get all available gradient parameter names

Returns a list of parameter names for which gradients are currently available.

pub fn get_fusion_stats(&self) -> FusionStats

Get fusion statistics