Struct juice::weight::WeightConfig

pub struct WeightConfig {
    pub name: String,
    pub share_mode: DimCheckMode,
    pub lr_mult: Option<f32>,
    pub decay_mult: Option<f32>,
    pub filler: Option<FillerType>,
}

Specifies training configuration for a weight blob.

Fields

name: String

The name of the weight blob – useful for sharing weights among layers, but never required otherwise. To share a weight between two layers, give it a (non-empty) name.

Default: “”

share_mode: DimCheckMode

Whether to require shared weights to have the same shape, or just the same count

Default: DimCheckMode::Strict

lr_mult: Option<f32>

The multiplier on the global learning rate for this parameter.

Default: 1.0f32

decay_mult: Option<f32>

The multiplier on the global weight decay for this parameter.

Default: 1.0f32

filler: Option<FillerType>

The filler that initializes the weights in the weight blob.

Default: None

Implementations

impl WeightConfig

pub fn check_dimensions<T>( &self, tensor_one: &SharedTensor<T>, tensor_two: &SharedTensor<T>, param_name: String, owner_name: String, layer_name: String ) -> Result<(), String>

Checks dimensions of two blobs according to the share_mode. Returns an error if there is a count/shape mismatch.

pub fn lr_mult(&self) -> f32

The multiplier on the global learning rate for this weight blob.

pub fn decay_mult(&self) -> f32

The multiplier on the global weight decay for this weight blob.

Trait Implementations

impl Clone for WeightConfig

fn clone(&self) -> WeightConfig

Returns a copy of the value.

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

Performs copy-assignment from source.

impl Debug for WeightConfig

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

Formats the value using the given formatter.

impl Default for WeightConfig

fn default() -> WeightConfig

Returns the “default value” for a type.