Enum Activations 

pub enum Activations {
    Exact,
    Int8,
    Int8TwoPass,
}

How the reference’s projections by quantized weights read their activations. Exact is the model’s arithmetic; the others are the activations a quantized matmul reads under llama.cpp’s rule — a quantized weight reads a quantized activation — computed as the GPU lane’s kernels compute them, so a lane can be held to its own precision.

Variants

Exact

As computed, in f64.

Int8

Rounded to f32, then each 32 values one int8 word over a shared scale max|v| / 127 (1 for a zero block), rounded to the nearest, ties to even.

Int8TwoPass

The high word as Int8 (the scale’s floor 1e-30, clamped to −128..127) and a low int8 word holding the remainder v − hi·s at a 128th of the scale.

Trait Implementations

impl Clone for Activations

fn clone(&self) -> Activations

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for Activations

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

Formats the value using the given formatter.

impl Default for Activations

fn default() -> Activations

Returns the “default value” for a type.

impl PartialEq for Activations

fn eq(&self, other: &Activations) -> 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 Activations

impl Eq for Activations

impl StructuralPartialEq for Activations