Struct TrainerBuilder 

pub struct TrainerBuilder { /* private fields */ }

Builder for TrainerConfig.

Implementations

impl TrainerBuilder

pub fn new() -> Self

Creates a builder with TrainerConfig::default settings.

pub fn target_vocab_size(self, value: usize) -> Self

Sets the desired vocabulary size (including base byte tokens).

pub fn min_frequency(self, value: usize) -> Self

Sets the minimum merge frequency.

pub fn allowed_token_lengths<I>(self, lengths: I) -> Self

where I: IntoIterator<Item = usize>,

Overrides the allowed token lengths.

pub fn show_progress(self, enabled: bool) -> Self

Enables or disables per-iteration logging.

pub fn special_tokens<I, S>(self, tokens: I) -> Self

where I: IntoIterator<Item = S>, S: Into<String>,

Appends custom special tokens after the built-in required tokens.

pub fn append_special_tokens<I, S>(self, tokens: I) -> Self

where I: IntoIterator<Item = S>, S: Into<String>,

Appends additional special tokens after the reasoning set (preserves order, deduplicated in build).

pub fn plateau_thresholds( self, floor: usize, patience: usize, divisor: usize, ) -> Self

Configures plateau-based early stopping thresholds.

pub fn plateau_stop_enabled(self, enabled: bool) -> Self

Enables plateau-based early stopping.

pub fn max_merge_iterations(self, value: Option<usize>) -> Self

Sets a hard merge iteration limit.

pub fn preprocessor(self, config: PreprocessorConfig) -> Self

Configures the preprocessing mode applied before training.

pub fn require_letter_whitespace_merges(self, enabled: bool) -> Self

Enforces that whitespace-containing merges include ASCII letters when enabled.

pub fn forbid_leading_whitespace_merges(self, enabled: bool) -> Self

Forbids merges that would introduce leading ASCII whitespace unless the token is pure whitespace.

pub fn reasoning_tokens_enabled(self, enabled: bool) -> Self

Enables or disables the optional reasoning/argument special tokens.

pub fn preprocessor_split_probability(self, probability: f64) -> Self

Overrides only the preprocessor split probability.

pub fn preprocessor_seed(self, seed: Option<u64>) -> Self

Sets an optional RNG seed for probabilistic preprocessing.

pub fn build(self) -> Result<TrainerConfig>

Finalises the builder, returning a validated TrainerConfig.

Trait Implementations

impl Clone for TrainerBuilder

fn clone(&self) -> TrainerBuilder

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TrainerBuilder

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

Formats the value using the given formatter.

impl Default for TrainerBuilder

fn default() -> Self

Returns the “default value” for a type.