Struct LogitAdjustment 

pub struct LogitAdjustment { /* private fields */ }

A vector subtracted from target logits to steer away from unsafe output. Sparse and integer (milli-logits) for deterministic, allocation-light steps.

Implementations

impl LogitAdjustment

pub fn none() -> LogitAdjustment

pub fn with_penalties(penalties: Vec<(u32, i32)>) -> LogitAdjustment

pub fn is_empty(&self) -> bool

pub fn delta_for(&self, token: u32) -> i32

The milli-logit delta to add for token (0 if unaffected).

pub fn l1_norm_milli(&self) -> u32

L1 norm in milli-units — what LogitsSteered.adjustment_norm_milli reports to telemetry.

Trait Implementations

impl Clone for LogitAdjustment

fn clone(&self) -> LogitAdjustment

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for LogitAdjustment

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

Formats the value using the given formatter.

impl Default for LogitAdjustment

fn default() -> LogitAdjustment

Returns the “default value” for a type.

impl Eq for LogitAdjustment

impl PartialEq for LogitAdjustment

fn eq(&self, other: &LogitAdjustment) -> 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 StructuralPartialEq for LogitAdjustment