Struct QuantConfig 

pub struct QuantConfig {
    pub dtype: DType,
    pub scheme: QScheme,
    pub enable_fake_quant: bool,
    pub observer_type: ObserverType,
    pub backend: QuantBackend,
    pub reduce_range: ReduceRange,
    pub qint_min: Option<i32>,
    pub qint_max: Option<i32>,
    pub eps: f32,
    pub averaging_constant: f32,
    pub ch_axis: Option<usize>,
    pub group_size: Option<usize>,
}

Quantization configuration

Fields

dtype: DType

scheme: QScheme

enable_fake_quant: bool

observer_type: ObserverType

backend: QuantBackend

reduce_range: ReduceRange

qint_min: Option<i32>

qint_max: Option<i32>

eps: f32

averaging_constant: f32

ch_axis: Option<usize>

group_size: Option<usize>

Group size for group-wise quantization

Implementations

impl QuantConfig

pub fn new() -> Self

Create a new quantization config with defaults

pub fn int8() -> Self

Create config for INT8 quantization

pub fn int4() -> Self

Create config for INT4 quantization

pub fn binary() -> Self

Create config for binary quantization

pub fn ternary() -> Self

Create config for ternary quantization

pub fn mixed_precision() -> Self

Create config for mixed precision

pub fn uint8() -> Self

Create config for UINT8 quantization

pub fn per_channel(ch_axis: usize) -> Self

Create config for per-channel quantization

pub fn group_wise(ch_axis: usize, group_size: usize) -> Self

Create config for group-wise quantization

pub fn qat() -> Self

Create config for QAT (Quantization Aware Training)

pub fn with_backend(self, backend: QuantBackend) -> Self

Create config with specific backend

pub fn with_observer(self, observer_type: ObserverType) -> Self

Set observer type

pub fn with_scheme(self, scheme: QScheme) -> Self

Set quantization scheme

pub fn with_fake_quant(self, enable: bool) -> Self

Enable/disable fake quantization

pub fn with_reduce_range(self, reduce_range: ReduceRange) -> Self

Set reduce range option

pub fn with_group_size(self, group_size: usize) -> Self

Set group size for group-wise quantization

pub fn get_qint_range(&self) -> (i32, i32)

Get effective quantization range considering scheme and reduce_range

pub fn validate(&self) -> TorshResult<()>

Validate configuration

Trait Implementations

impl Clone for QuantConfig

fn clone(&self) -> QuantConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for QuantConfig

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

Formats the value using the given formatter.

impl Default for QuantConfig

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for QuantConfig

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for QuantConfig

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

where __S: Serializer,

Serialize this value into the given Serde serializer.