Mila::Dnn::LpeConfig Class Reference

Configuration class for the Learned Positional Encoder. More...

Inheritance diagram for Mila::Dnn::LpeConfig:

Collaboration diagram for Mila::Dnn::LpeConfig:

Public Member Functions
void	fromMetadata (const SerializationMetadata &meta) override
	Populate configuration from serialization metadata.
size_t	getEmbeddingDim () const
	Get the configured embedding dimension.
size_t	getMaxSequenceLength () const
	Get the configured maximum sequence length.
size_t	getVocabularyLength () const
	Get the configured vocabulary length.
SerializationMetadata	toMetadata () const override
	Convert configuration to serialization metadata.
std::string	toString () const override
	Produce a short, human-readable summary of this configuration.
void	validate () const override
	Validate configuration parameters.
template<typename Self>
decltype(auto)	withEmbeddingDim (this Self &&self, size_t embedding_dim)
	C++23-style fluent setter for embedding dimension.
template<typename Self>
decltype(auto)	withMaxSequenceLength (this Self &&self, size_t max_seq_len)
	C++23-style fluent setter for maximum sequence length.
template<typename Self>
decltype(auto)	withVocabularyLength (this Self &&self, size_t vocab_len)
	C++23-style fluent setter for vocabulary length.
Public Member Functions inherited from Mila::Dnn::ComponentConfig
virtual	~ComponentConfig ()=default
	Virtual destructor for polymorphic base.

Private Attributes
size_t	embedding_dim_ = 0
	The embedding dimension size.
size_t	max_seq_len_ = 512
	The maximum sequence length.
size_t	vocab_len_ = 50000
	The vocabulary size.

Detailed Description

Configuration class for the Learned Positional Encoder.

Provides a type-safe fluent interface for configuring Encoder.

Member Function Documentation

fromMetadata()

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

inlineoverridevirtual

Populate configuration from serialization metadata.

Reads available fields from the provided metadata and updates the configuration object accordingly.

Implements Mila::Dnn::ComponentConfig.

Here is the call graph for this function:

getEmbeddingDim()

size_t Mila::Dnn::LpeConfig::getEmbeddingDim ( ) const

inline

Get the configured embedding dimension.

Returns

size_t The embedding dimension

getMaxSequenceLength()

size_t Mila::Dnn::LpeConfig::getMaxSequenceLength ( ) const

inline

Get the configured maximum sequence length.

Returns

size_t The maximum sequence length

getVocabularyLength()

size_t Mila::Dnn::LpeConfig::getVocabularyLength ( ) const

inline

Get the configured vocabulary length.

Returns

size_t The vocabulary length

toMetadata()

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

inlineoverridevirtual

Convert configuration to serialization metadata.

Produces a SerializationMetadata object containing the configuration fields suitable for writing into an archive by the caller.

Implements Mila::Dnn::ComponentConfig.

Here is the call graph for this function:

toString()

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

inlineoverridevirtual

Produce a short, human-readable summary of this configuration.

Overrides ComponentConfig::toString() to include Encoder-specific fields.

Implements Mila::Dnn::ComponentConfig.

validate()

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

inlineoverridevirtual

Validate configuration parameters.

Exceptions

std::invalid_argument

If validation fails

Implements Mila::Dnn::ComponentConfig.

Here is the caller graph for this function:

withEmbeddingDim()

template<typename Self>

decltype(auto) Mila::Dnn::LpeConfig::withEmbeddingDim ( this Self && self, size_t embedding_dim )

inline

C++23-style fluent setter for embedding dimension.

withMaxSequenceLength()

template<typename Self>

decltype(auto) Mila::Dnn::LpeConfig::withMaxSequenceLength ( this Self && self, size_t max_seq_len )

inline

C++23-style fluent setter for maximum sequence length.

withVocabularyLength()

template<typename Self>

decltype(auto) Mila::Dnn::LpeConfig::withVocabularyLength ( this Self && self, size_t vocab_len )

inline

C++23-style fluent setter for vocabulary length.

Here is the caller graph for this function:

Member Data Documentation

embedding_dim_

size_t Mila::Dnn::LpeConfig::embedding_dim_ = 0

private

The embedding dimension size.

max_seq_len_

size_t Mila::Dnn::LpeConfig::max_seq_len_ = 512

private

The maximum sequence length.

vocab_len_

size_t Mila::Dnn::LpeConfig::vocab_len_ = 50000

private

The vocabulary size.

---

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

• /__w/Mila/Mila/Mila/Src/Dnn/Components/Encodings/Lpe/Lpe.Config.ixx