Struct AdamWConfig 

pub struct AdamWConfig { /* private fields */ }

AdamW Configuration.

Implementations

impl AdamWConfig

pub fn new() -> AdamWConfig

Create a new instance of the config.

Arguments

Optional Arguments

grad_clipping

Gradient Clipping config.

• Defaults to None

Default Arguments

beta_1

Parameter for AdamW.

• Defaults to 0.9

beta_2

Parameter for AdamW.

• Defaults to 0.999

epsilon

A value required for numerical stability.

• Defaults to 1e-5

weight_decay

Weight decay config.

• Defaults to 1e-4

cautious_weight_decay

Cautious weight decay config.

See: https://arxiv.org/abs/2510.12402

• Defaults to false

amsgrad

Whether to use AMSGrad algorithm

• Defaults to false

impl AdamWConfig

pub fn with_beta_1(self, beta_1: f32) -> AdamWConfig

Sets the value for the field beta_1.

Parameter for AdamW.

• Defaults to 0.9

pub fn with_beta_2(self, beta_2: f32) -> AdamWConfig

Sets the value for the field beta_2.

Parameter for AdamW.

• Defaults to 0.999

pub fn with_epsilon(self, epsilon: f32) -> AdamWConfig

Sets the value for the field epsilon.

A value required for numerical stability.

• Defaults to 1e-5

pub fn with_weight_decay(self, weight_decay: f32) -> AdamWConfig

Sets the value for the field weight_decay.

Weight decay config.

• Defaults to 1e-4

pub fn with_cautious_weight_decay( self, cautious_weight_decay: bool, ) -> AdamWConfig

Sets the value for the field cautious_weight_decay.

Cautious weight decay config.

See: https://arxiv.org/abs/2510.12402

• Defaults to false

pub fn with_amsgrad(self, amsgrad: bool) -> AdamWConfig

Sets the value for the field amsgrad.

Whether to use AMSGrad algorithm

• Defaults to false

pub fn with_grad_clipping( self, grad_clipping: Option<GradientClippingConfig>, ) -> AdamWConfig

Sets the value for the field grad_clipping.

Gradient Clipping config.

• Defaults to None

impl AdamWConfig

pub fn build(&self) -> AdamW

Build an AdamW from the config.

The bare optimizer, which ModuleOptimizer::with_group takes to optimize one parameter group. init is the whole-module counterpart, and the only one that applies the configured gradient clipping.

pub fn init(&self) -> ModuleOptimizer

Initialize AdamW optimizer.

Returns

Returns an optimizer that can be used to optimize a module.

Trait Implementations

impl Clone for AdamWConfig

fn clone(&self) -> AdamWConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Config for AdamWConfig

fn save<P>(&self, file: P) -> Result<(), Error>

where P: AsRef<Path>,

Saves the configuration to a file.

fn load<P>(file: P) -> Result<Self, ConfigError>

where P: AsRef<Path>,

Loads the configuration from a file.

fn load_binary(data: &[u8]) -> Result<Self, ConfigError>

Loads the configuration from a binary buffer.

impl Debug for AdamWConfig

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for AdamWConfig

fn deserialize<D>( deserializer: D, ) -> Result<AdamWConfig, <D as Deserializer<'de>>::Error>

where D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Display for AdamWConfig

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

Formats the value using the given formatter.

impl Serialize for AdamWConfig

fn serialize<S>( &self, serializer: S, ) -> Result<<S as Serializer>::Ok, <S as Serializer>::Error>

where S: Serializer,

Serialize this value into the given Serde serializer.