Enum PruningStrategy 

pub enum PruningStrategy {
    Magnitude {
        threshold_percentile: u8,
    },
    BlockWise {
        block_size: (usize, usize),
    },
    ChannelWise {
        channels_to_prune: usize,
    },
    AttentionHead {
        heads_to_prune: usize,
    },
    Movement {
        sensitivity: f32,
    },
    GradualMagnitude {
        initial_sparsity: f32,
        final_sparsity: f32,
    },
}

Pruning strategies for tensor compression

Variants

Magnitude

Magnitude-based pruning: remove weights below threshold

Fields

threshold_percentile: u8

BlockWise

Structured pruning: remove entire blocks

Fields

block_size: (usize, usize)

ChannelWise

Channel-wise pruning for convolutional layers

Fields

channels_to_prune: usize

AttentionHead

Attention head pruning for transformer models

Fields

heads_to_prune: usize

Movement

Movement pruning: prune based on weight movement during training

Fields

sensitivity: f32

GradualMagnitude

Gradual magnitude pruning with schedule

Fields

initial_sparsity: f32

final_sparsity: f32

Implementations

impl PruningStrategy

pub fn expected_sparsity(&self) -> f32

Get the expected sparsity for this strategy

pub fn is_structured(&self) -> bool

Check if this is a structured pruning strategy

Trait Implementations

impl Clone for PruningStrategy

fn clone(&self) -> PruningStrategy

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for PruningStrategy

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

Formats the value using the given formatter.

impl PartialEq for PruningStrategy

fn eq(&self, other: &PruningStrategy) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Copy for PruningStrategy

impl StructuralPartialEq for PruningStrategy