Struct WeightedAccumulationState 

pub struct WeightedAccumulationState {
    pub dtype: FloatDType,
    pub loss_scale: f64,
    pub total_weight: f64,
    pub microbatches: u64,
}

Explicit normalization and continuation state of an accumulation window.

Fields

dtype: FloatDType

F32/F64 arithmetic used before adding each microbatch’s gradients.

loss_scale: f64

Loss multiplier actually applied by the caller throughout this window.

total_weight: f64

Sum of caller-supplied effective token/sample weights, not batch count.

microbatches: u64

Number of accepted microbatches, including explicit zero-weight batches.

Trait Implementations

impl Clone for WeightedAccumulationState

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for WeightedAccumulationState

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for WeightedAccumulationState

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl PartialEq for WeightedAccumulationState

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

Equality operator ==.

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

Inequality operator !=.

impl<B: Backend> Record<B> for WeightedAccumulationState

type Item<S: PrecisionSettings> = WeightedAccumulationState

Type of the item that can be serialized and deserialized.

fn into_item<S: PrecisionSettings>(self) -> Self

Convert the current record into the corresponding item that follows the given settings.

fn from_item<S: PrecisionSettings>(item: Self, _device: &B::Device) -> Self

Convert the given item into a record.

impl Serialize for WeightedAccumulationState

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

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for WeightedAccumulationState