Struct CrossEntropyLossConfig 

pub struct CrossEntropyLossConfig {
    pub pad_tokens: Option<Vec<usize>>,
    pub weights: Option<Vec<f32>>,
    pub smoothing: Option<f32>,
    pub logits: bool,
}

Configuration to create a Cross-entropy loss using the init function.

Fields

pad_tokens: Option<Vec<usize>>

Create padded cross entropy.

Prevents pad tokens from impacting loss calculation.

An entirely padded batch returns NaN, since the mean has no valid samples.

weights: Option<Vec<f32>>

Create weighted cross-entropy.

The loss of a specific sample will simply be given by: weight * log(p(x)) * 1,

Pre-conditions

• The order of the weight vector should correspond to the label integer assignment.
• Targets assigned negative Int’s will not be allowed.

smoothing: Option<f32>

Create cross-entropy with label smoothing.

Hard labels {0, 1} will be changed to y_smoothed = y(1 - a) + a / nr_classes. Alpha = 0 would be the same as default.

logits: bool

Create cross-entropy with probabilities as input instead of logits.

Implementations

impl CrossEntropyLossConfig

pub fn new() -> CrossEntropyLossConfig

Create a new instance of the config.

Arguments

Optional Arguments

pad_tokens

Create padded cross entropy.

Prevents pad tokens from impacting loss calculation.

An entirely padded batch returns NaN, since the mean has no valid samples.

• Defaults to None

weights

Create weighted cross-entropy.

The loss of a specific sample will simply be given by: weight * log(p(x)) * 1,

Pre-conditions

• The order of the weight vector should correspond to the label integer assignment.
• Targets assigned negative Int’s will not be allowed.
• Defaults to None

smoothing

Create cross-entropy with label smoothing.

Hard labels {0, 1} will be changed to y_smoothed = y(1 - a) + a / nr_classes. Alpha = 0 would be the same as default.

• Defaults to None

Default Arguments

logits

Create cross-entropy with probabilities as input instead of logits.

• Defaults to true

impl CrossEntropyLossConfig

pub fn with_logits(self, logits: bool) -> CrossEntropyLossConfig

Sets the value for the field logits.

Create cross-entropy with probabilities as input instead of logits.

• Defaults to true

pub fn with_pad_tokens( self, pad_tokens: Option<Vec<usize>>, ) -> CrossEntropyLossConfig

Sets the value for the field pad_tokens.

Create padded cross entropy.

Prevents pad tokens from impacting loss calculation.

An entirely padded batch returns NaN, since the mean has no valid samples.

• Defaults to None

pub fn with_weights(self, weights: Option<Vec<f32>>) -> CrossEntropyLossConfig

Sets the value for the field weights.

Create weighted cross-entropy.

The loss of a specific sample will simply be given by: weight * log(p(x)) * 1,

Pre-conditions

• The order of the weight vector should correspond to the label integer assignment.
• Targets assigned negative Int’s will not be allowed.
• Defaults to None

pub fn with_smoothing(self, smoothing: Option<f32>) -> CrossEntropyLossConfig

Sets the value for the field smoothing.

Create cross-entropy with label smoothing.

Hard labels {0, 1} will be changed to y_smoothed = y(1 - a) + a / nr_classes. Alpha = 0 would be the same as default.

• Defaults to None

impl CrossEntropyLossConfig

pub fn init(&self, device: &Device) -> CrossEntropyLoss

Initialize Cross-entropy loss.

Trait Implementations

impl Clone for CrossEntropyLossConfig

fn clone(&self) -> CrossEntropyLossConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Config for CrossEntropyLossConfig

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 CrossEntropyLossConfig

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for CrossEntropyLossConfig

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

where D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Display for CrossEntropyLossConfig

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

Formats the value using the given formatter.

impl Serialize for CrossEntropyLossConfig

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.