Struct tokenizers::models::bpe::trainer::BpeTrainer

#[non_exhaustive]
pub struct BpeTrainer {
    pub min_frequency: u64,
    pub vocab_size: usize,
    pub show_progress: bool,
    pub special_tokens: Vec<AddedToken>,
    pub limit_alphabet: Option<usize>,
    pub initial_alphabet: HashSet<char>,
    pub continuing_subword_prefix: Option<String>,
    pub end_of_word_suffix: Option<String>,
    pub max_token_length: Option<usize>,
    /* private fields */
}

In charge of training a BPE model

Examples

use tokenizers::tokenizer::Trainer;
use tokenizers::models::bpe::{BPE, BpeTrainer};

let sequences = vec![ "Hello", "World" ];

let mut trainer = BpeTrainer::default();
trainer.feed(sequences.iter(), |s| Ok(vec![s.to_owned()]));

let mut model = BPE::default();
let special_tokens = trainer.train(&mut model).unwrap();

Fields (Non-exhaustive)

Non-exhaustive structs could have additional fields added in future. Therefore, non-exhaustive structs cannot be constructed in external crates using the traditional Struct { .. } syntax; cannot be matched against without a wildcard ..; and struct update syntax will not work.

min_frequency: u64

The minimum frequency a pair must have to produce a merge operation

vocab_size: usize

The target vocabulary size

show_progress: bool

Whether to show progress while training

special_tokens: Vec<AddedToken>

A list of special tokens that the model should know of

limit_alphabet: Option<usize>

Whether to limit the number of initial tokens that can be kept before computing merges

initial_alphabet: HashSet<char>

The initial alphabet we want absolutely to include. This allows to cover some characters that are not necessarily in the training set

continuing_subword_prefix: Option<String>

An optional prefix to use on any subword that exist only behind another one

end_of_word_suffix: Option<String>

An optional suffix to caracterize and end-of-word subword

max_token_length: Option<usize>

An optional parameter to limit the max length of any single token

Implementations

impl BpeTrainer

pub fn new(min_frequency: u64, vocab_size: usize) -> Self

pub fn builder() -> BpeTrainerBuilder

pub fn do_train( &self, word_counts: &HashMap<String, u64>, model: &mut BPE ) -> Result<Vec<AddedToken>>

Trait Implementations

impl Clone for BpeTrainer

fn clone(&self) -> BpeTrainer

Returns a copy of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for BpeTrainer

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for BpeTrainer

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for BpeTrainer

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl From<BpeTrainer> for TrainerWrapper

fn from(from: BpeTrainer) -> Self

Converts to this type from the input type.

impl PartialEq for BpeTrainer

fn eq(&self, other: &BpeTrainer) -> bool

This method tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

This method tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Serialize for BpeTrainer

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl Trainer for BpeTrainer

fn train(&self, model: &mut BPE) -> Result<Vec<AddedToken>>

Train a BPE model

fn should_show_progress(&self) -> bool

Whether we should show progress

type Model = BPE

fn feed<I, S, F>(&mut self, iterator: I, process: F) -> Result<()>

where I: Iterator<Item = S> + Send, S: AsRef<str> + Send, F: Fn(&str) -> Result<Vec<String>> + Sync,

Process an iterator of sequences, calling process for each of them in order to pre-process the said sequence as relevant.

impl Eq for BpeTrainer

impl StructuralPartialEq for BpeTrainer