Struct MixedPrecisionConfig 

pub struct MixedPrecisionConfig {
    pub init_scale: f32,
    pub growth_factor: f32,
    pub backoff_factor: f32,
    pub growth_interval: u32,
    pub min_scale: f32,
    pub max_scale: f32,
    pub use_bfloat16: bool,
}

Configuration for dynamic loss scaling.

Fields

init_scale: f32

Initial loss scale factor.

growth_factor: f32

Multiplier applied when growing the scale.

backoff_factor: f32

Multiplier applied when an overflow is observed.

growth_interval: u32

Number of consecutive overflow-free steps before growing.

min_scale: f32

Lower clamp for the scale.

max_scale: f32

Upper clamp for the scale.

use_bfloat16: bool

Use bfloat16 rather than float16 for the reduced-precision copy.

Trait Implementations

impl Clone for MixedPrecisionConfig

fn clone(&self) -> MixedPrecisionConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for MixedPrecisionConfig

impl Debug for MixedPrecisionConfig

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

Formats the value using the given formatter.

impl Default for MixedPrecisionConfig

fn default() -> Self

Returns the “default value” for a type.

impl PartialEq for MixedPrecisionConfig

fn eq(&self, other: &MixedPrecisionConfig) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for MixedPrecisionConfig