Mila::Dnn::LinearConfig Class Reference

Configuration object for a Linear (fully connected) layer. More...

Inheritance diagram for Mila::Dnn::LinearConfig:

Collaboration diagram for Mila::Dnn::LinearConfig:

Public Member Functions
	LinearConfig (dim_t input_features, dim_t output_features)
	Construct a LinearConfig with required feature dimensions.
void	fromMetadata (const SerializationMetadata &meta) override
	Populate configuration from provided metadata.
dim_t	getInputFeatures () const noexcept
	Get the configured number of input features.
dim_t	getOutputFeatures () const noexcept
	Get the configured number of output features.
bool	hasBias () const noexcept
	Query whether the bias term is enabled.
SerializationMetadata	toMetadata () const override
	Convert configuration into SerializationMetadata.
std::string	toString () const override
	Human-readable summary suitable for logging.
void	validate () const override
	Validate the configuration values.
template<typename Self>
Self &&	withBias (this Self &&self, bool has_bias)
	Fluent setter for bias enable flag.
template<typename Self>
Self &&	withInputFeatures (this Self &&self, dim_t input_features)
	Fluent setter for input features.
template<typename Self>
Self &&	withOutputFeatures (this Self &&self, dim_t output_features)
	Fluent setter for output features.
Public Member Functions inherited from Mila::Dnn::ComponentConfig
virtual	~ComponentConfig ()=default
	Virtual destructor for polymorphic base.

Private Attributes
bool	has_bias_ { true }
	Whether the layer has a bias term.
dim_t	input_features_
	Number of input features (must be > 0).
dim_t	output_features_
	Number of output features (must be > 0).

Detailed Description

Configuration object for a Linear (fully connected) layer.

LinearConfig describes parameters required to construct a Linear layer: input and output feature dimensions and whether a bias is present.

Instances are lightweight value objects intended to be passed into module factories or constructors. Call validate() prior to constructing runtime objects to surface configuration errors early.

Constructor & Destructor Documentation

LinearConfig()

Mila::Dnn::LinearConfig::LinearConfig ( dim_t input_features, dim_t output_features )

inline

Construct a LinearConfig with required feature dimensions.

Parameters

input_features	Number of input features (must be > 0).
output_features	Number of output features (must be > 0).

Member Function Documentation

fromMetadata()

void Mila::Dnn::LinearConfig::fromMetadata ( const SerializationMetadata & meta )

inlineoverridevirtual

Populate configuration from provided metadata.

Missing keys are ignored, leaving defaults intact. Type mismatches result in no assignment (use tryGet* helpers).

Parameters

meta	Metadata to read configuration values from.

Implements Mila::Dnn::ComponentConfig.

Here is the call graph for this function:

getInputFeatures()

dim_t Mila::Dnn::LinearConfig::getInputFeatures ( ) const

inlinenoexcept

Get the configured number of input features.

Returns

dim_t Number of input features configured.

getOutputFeatures()

dim_t Mila::Dnn::LinearConfig::getOutputFeatures ( ) const

inlinenoexcept

Get the configured number of output features.

Returns

dim_t Number of output features configured.

hasBias()

bool Mila::Dnn::LinearConfig::hasBias ( ) const

inlinenoexcept

Query whether the bias term is enabled.

Returns

bool True if bias is enabled; false otherwise.

toMetadata()

SerializationMetadata Mila::Dnn::LinearConfig::toMetadata ( ) const

inlineoverridevirtual

Convert configuration into SerializationMetadata.

Produces keys:

• "precision" : integer (ComputePrecision::Policy)
• "input_features" : integer
• "output_features" : integer
• "has_bias" : boolean

Returns

SerializationMetadata Metadata representing this configuration.

Implements Mila::Dnn::ComponentConfig.

Here is the call graph for this function:

toString()

std::string Mila::Dnn::LinearConfig::toString ( ) const

inlineoverridevirtual

Human-readable summary suitable for logging.

Returns

std::string Compact description of the configuration.

Implements Mila::Dnn::ComponentConfig.

validate()

void Mila::Dnn::LinearConfig::validate ( ) const

inlineoverridevirtual

Validate the configuration values.

Throws std::invalid_argument when the configuration is invalid.

Implements Mila::Dnn::ComponentConfig.

withBias()

template<typename Self>

Self && Mila::Dnn::LinearConfig::withBias ( this Self && self, bool has_bias )

inline

Fluent setter for bias enable flag.

Template Parameters

Self	Concrete config type (deduced via explicit object parameter)

Parameters

has_bias

True to include a bias parameter, false to omit it.

Returns

Self&& Reference to this configuration for chaining.

Here is the caller graph for this function:

withInputFeatures()

template<typename Self>

Self && Mila::Dnn::LinearConfig::withInputFeatures ( this Self && self, dim_t input_features )

inline

Fluent setter for input features.

Parameters

input_features

Number of input features.

Returns

Self&& Reference to this configuration for chaining.

withOutputFeatures()

template<typename Self>

Self && Mila::Dnn::LinearConfig::withOutputFeatures ( this Self && self, dim_t output_features )

inline

Fluent setter for output features.

Parameters

output_features

Number of output features.

Returns

Self&& Reference to this configuration for chaining.

Member Data Documentation

has_bias_

bool Mila::Dnn::LinearConfig::has_bias_ { true }

private

Whether the layer has a bias term.

Default is true.

input_features_

dim_t Mila::Dnn::LinearConfig::input_features_

private

Number of input features (must be > 0).

output_features_

dim_t Mila::Dnn::LinearConfig::output_features_

private

Number of output features (must be > 0).

---

The documentation for this class was generated from the following file:

• /__w/Mila/Mila/Mila/Src/Dnn/Components/Linear/LinearConfig.ixx