Struct SensitivityAnalyzer 

pub struct SensitivityAnalyzer;

Analyses per-layer quantization sensitivity.

Implementations

impl SensitivityAnalyzer

pub fn new() -> Self

Create a new sensitivity analyser.

pub fn analyze_layer( &self, weights: &[f32], bits_range: &[u32], name: impl Into<String>, ) -> QuantResult<LayerSensitivity>

Compute quantization sensitivity for one layer across bits_range.

Parameters

• weights — flat weight tensor (any layout).
• bits_range — candidate bit-widths (e.g., &[2, 3, 4, 8]).
• name — optional layer label.

Errors

• QuantError::EmptyInput — weights is empty.
• QuantError::InvalidBitWidth — any bit-width in bits_range is 0 or > 16.

pub fn analyze_multiple<'a>( &self, layers: &[(&'a str, &'a [f32])], bits_range: &[u32], ) -> QuantResult<Vec<LayerSensitivity>>

Analyse multiple layers and return their sensitivity profiles.

Parameters

• layers — list of (name, weights) pairs.
• bits_range — candidate bit-widths to test for each layer.

Errors

Propagates errors from analyze_layer.

Trait Implementations

impl Clone for SensitivityAnalyzer

fn clone(&self) -> SensitivityAnalyzer

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SensitivityAnalyzer

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

Formats the value using the given formatter.

impl Default for SensitivityAnalyzer

fn default() -> SensitivityAnalyzer

Returns the “default value” for a type.