Struct PruningCallback 

pub struct PruningCallback { /* private fields */ }

Callback for applying pruning during training.

Integrates with the training loop to apply pruning at scheduled steps, collect calibration data for activation-weighted methods, and log pruning metrics.

Toyota Way: Kaizen (Continuous Improvement)

Gradual pruning allows the model to adapt incrementally to sparsity.

Example

use entrenar::prune::{PruningCallback, PruningConfig, PruningSchedule};

let config = PruningConfig::new()
    .with_schedule(PruningSchedule::Gradual {
        start_step: 1000,
        end_step: 5000,
        initial_sparsity: 0.0,
        final_sparsity: 0.5,
        frequency: 100,
    })
    .with_target_sparsity(0.5);

let callback = PruningCallback::new(config);
trainer.add_callback(callback);

Implementations

impl PruningCallback

pub fn new(config: PruningConfig) -> Self

Create a new pruning callback with the given configuration.

Arguments

• config - Pruning configuration

pub fn with_calibration( config: PruningConfig, cal_config: CalibrationConfig, ) -> Self

Create a pruning callback with custom calibration configuration.

pub fn set_enabled(&mut self, enabled: bool)

Enable or disable the callback.

pub fn is_enabled(&self) -> bool

Check if the callback is enabled.

pub fn current_sparsity(&self) -> f32

Get the current achieved sparsity.

pub fn target_sparsity(&self) -> f32

Get the target sparsity from the configuration.

pub fn parameters_pruned(&self) -> usize

Get the total number of parameters pruned.

pub fn schedule(&self) -> &PruningSchedule

Get the pruning schedule.

pub fn is_complete(&self) -> bool

Check if pruning is complete.

pub fn last_prune_step(&self) -> Option<usize>

Get the step at which pruning last occurred.

pub fn set_current_sparsity(&mut self, sparsity: f32)

Update current sparsity (for testing or manual updates).

pub fn config(&self) -> &PruningConfig

Get the configuration.

pub fn progress(&self) -> f32

Compute progress through the pruning schedule (0.0 to 1.0).

Trait Implementations

impl Clone for PruningCallback

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for PruningCallback

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

Formats the value using the given formatter.

impl TrainerCallback for PruningCallback

fn on_train_begin(&mut self, _ctx: &CallbackContext) -> CallbackAction

Called before training starts

fn on_step_end(&mut self, ctx: &CallbackContext) -> CallbackAction

Called after each training step

fn on_train_end(&mut self, _ctx: &CallbackContext)

Called after training ends

fn name(&self) -> &'static str

Get callback name for logging

fn on_epoch_begin(&mut self, _ctx: &CallbackContext) -> CallbackAction

Called before each epoch

fn on_epoch_end(&mut self, _ctx: &CallbackContext) -> CallbackAction

Called after each epoch

fn on_step_begin(&mut self, _ctx: &CallbackContext) -> CallbackAction

Called before each training step

fn on_validation(&mut self, _ctx: &CallbackContext) -> CallbackAction

Called when validation is performed