Struct Adam

pub struct Adam { /* private fields */ }

Adam Optimizer, with possibility of using AdaBound

Implementations

impl Adam

pub fn set_lr(&mut self, learning_rate: Float) -> &mut Self

Set learning rate

pub fn set_final_lr(&mut self, learning_rate: Float) -> &mut Self

Set final learning rate for AdaBound (SGD)

pub fn set_lambda(&mut self, coeff: Float) -> &mut Self

Set lambda factor for weight decay

pub fn set_gamma(&mut self, coeff: Float) -> &mut Self

Set gamma factor for AdaBound bounding convergence

pub fn set_beta1(&mut self, beta: Float) -> &mut Self

Set beta1 coefficient (for exponential moving average of first moment)

pub fn set_beta2(&mut self, beta: Float) -> &mut Self

Set beta2 coefficient (for exponential moving average of second moment)

pub fn set_eps(&mut self, epsilon: Float) -> &mut Self

Set epsilon to avoid divide by zero (fuzz factor)

pub fn set_adabound(&mut self, use_bound: bool) -> &mut Self

Set usage of AdaBound

pub fn to_json(&self) -> String

Encodes the optimizer as a JSON string.

pub fn from_json(encoded: &str) -> Adam

Builds a new optimizer from a JSON string.

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

Saves the model to a file

pub fn load(file: &str) -> Result<Adam, Error>

Creates a model from a previously saved file

Trait Implementations

impl Clone for Adam

fn clone(&self) -> Adam

Returns a copy of the value.

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

Performs copy-assignment from source.

impl Debug for Adam

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

Formats the value using the given formatter.

impl Default for Adam

fn default() -> Self

Create new Adam optimizer instance using default hyperparameters (lr = 0.001, lambda = 0, beta1 = 0.9, beta2 = 0.999, eps = 1e-8, adabound = false, final_lr = 0.1, gamma: 0.001) Also try higher LR; beta2 = 0.99; try adabound!

impl<'de> Deserialize<'de> for Adam

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Optimizer for Adam

fn get_delta(&mut self, params: &[Float], grad: &[Float]) -> Vec<Float>

Compute delta update from params and gradient

fn get_t(&self) -> usize

Retrieve the timestep (to allow computing manual learning rate decay)

impl Serialize for Adam

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

where __S: Serializer,

Serialize this value into the given Serde serializer.