Enum QuantizationLevel 

#[non_exhaustive]
pub enum QuantizationLevel {
    None,
    Float16,
    BFloat16,
    Int8,
    Int4,
}

Model weight quantisation levels.

Examples

use ai_hwaccel::QuantizationLevel;

let q = QuantizationLevel::Int8;
assert_eq!(q.bits_per_param(), 8);
assert!((q.memory_reduction_factor() - 4.0).abs() < f64::EPSILON);

Variants (Non-exhaustive)

Non-exhaustive enums could have additional variants added in future. Therefore, when matching against variants of non-exhaustive enums, an extra wildcard arm must be added to account for any future variants.

None

Full precision — FP32, 32 bits per parameter.

Float16

Half precision — FP16, 16 bits per parameter.

BFloat16

Brain floating point — BF16, 16 bits per parameter.

Int8

8-bit integer quantisation.

Int4

4-bit integer quantisation (GPTQ / AWQ style).

Implementations

impl QuantizationLevel

pub fn bits_per_param(&self) -> u32

Number of bits used per model parameter.

pub fn memory_reduction_factor(&self) -> f64

Memory reduction factor relative to FP32.

Trait Implementations

impl Clone for QuantizationLevel

fn clone(&self) -> QuantizationLevel

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for QuantizationLevel

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for QuantizationLevel

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Display for QuantizationLevel

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

Formats the value using the given formatter.

impl Hash for QuantizationLevel

fn hash<__H: Hasher>(&self, state: &mut __H)

Feeds this value into the given Hasher.

fn hash_slice<H>(data: &[Self], state: &mut H)

where H: Hasher, Self: Sized,

Feeds a slice of this type into the given Hasher.

impl PartialEq for QuantizationLevel

fn eq(&self, other: &QuantizationLevel) -> 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 Serialize for QuantizationLevel

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl TryFrom<u32> for QuantizationLevel

fn try_from(bits: u32) -> Result<Self, u32>

Convert from bit width to quantisation level.

• 32 → None (FP32)
• 16 → Float16
• 8 → Int8
• 4 → Int4

type Error = u32

The type returned in the event of a conversion error.

impl Copy for QuantizationLevel

impl Eq for QuantizationLevel

impl StructuralPartialEq for QuantizationLevel