Mila::Data::CharTrainer Class Reference

Character-level tokenizer trainer. More...

Collaboration diagram for Mila::Data::CharTrainer:

Public Member Functions
	CharTrainer (const CharVocabularyConfig &config=CharVocabularyConfig{})
	Construct with configuration.
void	addCorpusFromFile (const fs::path &path)
	Add corpus text from file.
void	addCorpusFromStream (std::istream &stream)
	Add corpus text from input stream.
void	clearCorpus ()
	Clear accumulated corpus.
const CharVocabularyConfig &	getConfig () const
	Get the trainer configuration.
size_t	getCorpusSize () const
	Get accumulated corpus size in bytes.
CharVocabulary	train ()
	Build vocabulary on accumulated corpus.

Private Attributes
CharVocabularyConfig	config_
std::string	corpus_

Detailed Description

Character-level tokenizer trainer.

Manages corpus accumulation and delegates vocabulary building to CharVocabulary factory methods. Provides a convenient API for incremental corpus loading and batch training.

Note: Character tokenization is simple enough that direct use of CharVocabulary::train() or CharVocabulary::trainFromFile() is often preferred. This trainer is maintained for API consistency with BpeTrainer and potential future extensions.

Constructor & Destructor Documentation

CharTrainer()

Mila::Data::CharTrainer::CharTrainer ( const CharVocabularyConfig & config = CharVocabularyConfig{} )

inlineexplicit

Construct with configuration.

Parameters

config

Character vocabulary configuration.

Exceptions

std::invalid_argument

if config is invalid.

Member Function Documentation

addCorpusFromFile()

void Mila::Data::CharTrainer::addCorpusFromFile ( const fs::path & path )

inline

Add corpus text from file.

Convenience method for loading corpus from filesystem.

Parameters

path	Path to corpus text file.

Exceptions

std::runtime_error

if file cannot be opened.

Here is the call graph for this function:

Here is the caller graph for this function:

addCorpusFromStream()

void Mila::Data::CharTrainer::addCorpusFromStream ( std::istream & stream )

inline

Add corpus text from input stream.

Accumulates text for training. Can be called multiple times to add corpus from different sources.

Parameters

stream

Input stream containing corpus text.

Here is the caller graph for this function:

clearCorpus()

void Mila::Data::CharTrainer::clearCorpus ( )

inline

Clear accumulated corpus.

Frees memory used by corpus accumulation.

getConfig()

const CharVocabularyConfig & Mila::Data::CharTrainer::getConfig ( ) const

inline

Get the trainer configuration.

Returns

const CharVocabularyConfig& Configuration reference.

getCorpusSize()

size_t Mila::Data::CharTrainer::getCorpusSize ( ) const

inline

Get accumulated corpus size in bytes.

Returns

Size of accumulated corpus.

train()

CharVocabulary Mila::Data::CharTrainer::train ( )

inline

Build vocabulary on accumulated corpus.

Delegates to CharVocabulary::train() factory method. Clears accumulated corpus after training to free memory.

Returns

Built CharVocabulary instance.

Exceptions

std::runtime_error	if corpus is empty.
std::invalid_argument	if config is invalid.

Here is the call graph for this function:

Here is the caller graph for this function:

Member Data Documentation

config_

CharVocabularyConfig Mila::Data::CharTrainer::config_

private

corpus_

std::string Mila::Data::CharTrainer::corpus_

private

---

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

• /__w/Mila/Mila/Mila/Src/Data/Tokenizers/Char/CharTrainer.ixx