Mila::Dnn::ComponentConfig Class Reference

Base configuration class for all neural network components. More...

Inheritance diagram for Mila::Dnn::ComponentConfig:

Public Member Functions
virtual	~ComponentConfig ()=default
	Virtual destructor to support proper polymorphic destruction.
const std::string &	getName () const
	Gets the configured component name.
ComputePrecision::Policy	getPrecision () const
	Gets the configured precision policy.
bool	isTraining () const
	Gets the configured training mode.
virtual void	validate () const
	Validates the configuration.
template<typename Self >
auto &	withName (this Self &&self, std::string name)
	Sets the name of the component with fluent interface.
template<typename Self >
auto &	withPrecision (this Self &&self, ComputePrecision::Policy policy)
	Sets the compute precision policy with fluent interface.
template<typename Self >
auto &	withTraining (this Self &&self, bool is_training)
	Sets the training mode with fluent interface.

Protected Attributes
bool	is_training_ = false
	Training mode flag, defaults to false (inference mode)
std::string	name_ = "unnamed"
	Component name, defaults to "unnamed" if not explicitly set.
ComputePrecision::Policy	precision_ = ComputePrecision::Policy::Auto
	Precision policy for computation, defaults to Auto.

Detailed Description

Base configuration class for all neural network components.

ComponentConfig serves as the foundation for all configuration classes in the Mila DNN framework. It provides common configuration properties like name, precision policy, and training mode that are shared across different neural network components.

Derived configuration classes should extend this base with component-specific configuration options and override validate() when needed.

See also

GeluConfig

LinearConfig

Constructor & Destructor Documentation

~ComponentConfig()

virtual Mila::Dnn::ComponentConfig::~ComponentConfig ( )

virtualdefault

Virtual destructor to support proper polymorphic destruction.

Member Function Documentation

getName()

const std::string & Mila::Dnn::ComponentConfig::getName ( ) const

inline

Gets the configured component name.

Returns

const std::string& The component name

Here is the caller graph for this function:

getPrecision()

ComputePrecision::Policy Mila::Dnn::ComponentConfig::getPrecision ( ) const

inline

Gets the configured precision policy.

Returns

ComputePrecision::Policy The precision policy

Here is the caller graph for this function:

isTraining()

bool Mila::Dnn::ComponentConfig::isTraining ( ) const

inline

Gets the configured training mode.

Returns

bool True if in training mode, false if in inference mode

validate()

virtual void Mila::Dnn::ComponentConfig::validate ( ) const

inlinevirtual

Validates the configuration.

Base implementation validates that the component name is not empty. Derived classes should call this base implementation and add their own validation logic.

Exceptions

std::invalid_argument

If the configuration is invalid

Reimplemented in Mila::Dnn::GeluConfig, Mila::Dnn::MLPConfig, Mila::Dnn::TransformerBlockConfig, Mila::Dnn::EncoderConfig, Mila::Dnn::LinearConfig, Mila::Dnn::MultiHeadAttentionConfig, Mila::Dnn::ResidualConfig, Mila::Dnn::SoftmaxConfig, Mila::Dnn::CrossEntropyConfig, Mila::Dnn::LayerNormConfig, and Mila::Dnn::DropoutConfig.

Here is the caller graph for this function:

withName()

template<typename Self >

auto & Mila::Dnn::ComponentConfig::withName ( this Self &&  self, std::string  name  )

inline

Sets the name of the component with fluent interface.

Parameters

name	The name to assign to this component

Returns

Reference to self for method chaining

Note

Names are validated during the validate() call, not at assignment time

Here is the caller graph for this function:

withPrecision()

template<typename Self >

auto & Mila::Dnn::ComponentConfig::withPrecision ( this Self &&  self, ComputePrecision::Policy  policy  )

inline

Sets the compute precision policy with fluent interface.

Parameters

policy

The compute precision policy to use

Returns

Reference to self for method chaining

withTraining()

template<typename Self >

auto & Mila::Dnn::ComponentConfig::withTraining ( this Self &&  self, bool  is_training  )

inline

Sets the training mode with fluent interface.

Parameters

is_training

True to put component in training mode, false for inference mode

Returns

Reference to self for method chaining

Member Data Documentation

is_training_

bool Mila::Dnn::ComponentConfig::is_training_ = false

protected

Training mode flag, defaults to false (inference mode)

name_

std::string Mila::Dnn::ComponentConfig::name_ = "unnamed"

protected

Component name, defaults to "unnamed" if not explicitly set.

precision_

ComputePrecision::Policy Mila::Dnn::ComponentConfig::precision_ = ComputePrecision::Policy::Auto

protected

Precision policy for computation, defaults to Auto.

---

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

• /home/runner/work/Mila/Mila/Mila/Src/Dnn/Common/ComponentConfig.ixx