Mila::Dnn::CrossEntropyConfig Class Reference

Configuration class for CrossEntropy module. More...

Inheritance diagram for Mila::Dnn::CrossEntropyConfig:

Collaboration diagram for Mila::Dnn::CrossEntropyConfig:

Public Member Functions
	CrossEntropyConfig (int64_t vocab_size)
	Constructor with required vocabulary size parameter.
const std::vector< float > &	getClassWeights () const
	Get the class weights.
float	getLabelSmoothing () const
	Get the label smoothing factor.
int64_t	getPaddingIndex () const
	Get the padding index.
bool	getReduction () const
	Check if loss should be reduced.
int64_t	getVocabSize () const
	Get the vocabulary size.
bool	ignorePadding () const
	Check if padding should be ignored.
void	validate () const
	Validate configuration parameters.
CrossEntropyConfig &	withClassWeights (const std::vector< float > &weights)
	Set class weights for weighted cross entropy.
CrossEntropyConfig &	withIgnorePadding (bool ignore_pad)
	Configure whether to ignore padding index.
CrossEntropyConfig &	withLabelSmoothing (float smoothing)
	Configure whether to apply label smoothing.
CrossEntropyConfig &	withPaddingIndex (int64_t pad_idx)
	Set the padding index to ignore.
CrossEntropyConfig &	withReduction (bool reduce)
	Configure whether to reduce the loss.
Public Member Functions inherited from Mila::Dnn::ComponentConfig
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.
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.

Private Attributes
std::vector< float >	class_weights_
bool	ignore_padding_ = false
float	label_smoothing_ = 0.0f
int64_t	padding_idx_ = -1
bool	reduce_ = true
int64_t	vocab_size_

Additional Inherited Members
Protected Attributes inherited from Mila::Dnn::ComponentConfig
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

Configuration class for CrossEntropy module.

Constructor & Destructor Documentation

CrossEntropyConfig()

Mila::Dnn::CrossEntropyConfig::CrossEntropyConfig ( int64_t  vocab_size )

inlineexplicit

Constructor with required vocabulary size parameter.

Parameters

vocab_size

The size of the vocabulary (number of possible classes)

Member Function Documentation

getClassWeights()

const std::vector< float > & Mila::Dnn::CrossEntropyConfig::getClassWeights ( ) const

inline

Get the class weights.

Here is the caller graph for this function:

getLabelSmoothing()

float Mila::Dnn::CrossEntropyConfig::getLabelSmoothing ( ) const

inline

Get the label smoothing factor.

Here is the caller graph for this function:

getPaddingIndex()

int64_t Mila::Dnn::CrossEntropyConfig::getPaddingIndex ( ) const

inline

Get the padding index.

Here is the caller graph for this function:

getReduction()

bool Mila::Dnn::CrossEntropyConfig::getReduction ( ) const

inline

Check if loss should be reduced.

Here is the caller graph for this function:

getVocabSize()

int64_t Mila::Dnn::CrossEntropyConfig::getVocabSize ( ) const

inline

Get the vocabulary size.

Here is the caller graph for this function:

ignorePadding()

bool Mila::Dnn::CrossEntropyConfig::ignorePadding ( ) const

inline

Check if padding should be ignored.

Here is the caller graph for this function:

validate()

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

inlinevirtual

Validate configuration parameters.

Exceptions

std::invalid_argument

If validation fails

Reimplemented from Mila::Dnn::ComponentConfig.

Here is the call graph for this function:

Here is the caller graph for this function:

withClassWeights()

CrossEntropyConfig & Mila::Dnn::CrossEntropyConfig::withClassWeights ( const std::vector< float > &  weights )

inline

Set class weights for weighted cross entropy.

Parameters

weights

Vector of weights for each class

Returns

CrossEntropyConfig& Reference to this for method chaining

withIgnorePadding()

CrossEntropyConfig & Mila::Dnn::CrossEntropyConfig::withIgnorePadding ( bool  ignore_pad )

inline

Configure whether to ignore padding index.

When true, targets with the specified padding index will not contribute to the loss.

Parameters

ignore_pad

Enable padding index ignoring

Returns

CrossEntropyConfig& Reference to this for method chaining

withLabelSmoothing()

CrossEntropyConfig & Mila::Dnn::CrossEntropyConfig::withLabelSmoothing ( float  smoothing )

inline

Configure whether to apply label smoothing.

Parameters

smoothing

Label smoothing factor (0.0 to 1.0)

Returns

CrossEntropyConfig& Reference to this for method chaining

withPaddingIndex()

CrossEntropyConfig & Mila::Dnn::CrossEntropyConfig::withPaddingIndex ( int64_t  pad_idx )

inline

Set the padding index to ignore.

Parameters

pad_idx

The padding index value to ignore in loss calculation

Returns

CrossEntropyConfig& Reference to this for method chaining

withReduction()

CrossEntropyConfig & Mila::Dnn::CrossEntropyConfig::withReduction ( bool  reduce )

inline

Configure whether to reduce the loss.

When true, returns the mean of losses. When false, returns per-sample losses.

Parameters

reduce

Whether to average the loss

Returns

CrossEntropyConfig& Reference to this for method chaining

Member Data Documentation

class_weights_

std::vector<float> Mila::Dnn::CrossEntropyConfig::class_weights_

private

ignore_padding_

bool Mila::Dnn::CrossEntropyConfig::ignore_padding_ = false

private

label_smoothing_

float Mila::Dnn::CrossEntropyConfig::label_smoothing_ = 0.0f

private

padding_idx_

int64_t Mila::Dnn::CrossEntropyConfig::padding_idx_ = -1

private

reduce_

bool Mila::Dnn::CrossEntropyConfig::reduce_ = true

private

vocab_size_

int64_t Mila::Dnn::CrossEntropyConfig::vocab_size_

private

---

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

• /home/runner/work/Mila/Mila/Mila/Src/Dnn/Modules/Losses/CrossEntropyConfig.ixx