Struct SamplerConfig 

pub struct SamplerConfig {
    pub seed: u64,
    pub batch_size: usize,
    pub ingestion_max_records: usize,
    pub chunking: ChunkingStrategy,
    pub recipes: Vec<TripletRecipe>,
    pub text_recipes: Vec<TextRecipe>,
    pub split: SplitRatios,
    pub allowed_splits: Vec<SplitLabel>,
}

Top-level sampler configuration.

Fields

seed: u64

RNG seed that controls deterministic sampling order.

batch_size: usize

Target number of samples per batch.

ingestion_max_records: usize

Max number of records kept in the ingestion cache for candidate sampling.

This is intentionally decoupled from batch_size so anchors/negatives can be drawn from a broader rolling pool.

Practical tuning: values above batch_size usually improve diversity and reduce short-horizon repetition; gains taper off as source/recipe/split constraints become the limiting factor. Higher values also increase memory.

For remote shard-backed sources (for example Hugging Face), larger initial targets may require fetching more shards before the first batch, so startup latency can increase based on shard sizes and network throughput.

chunking: ChunkingStrategy

Chunking behavior for long sections.

recipes: Vec<TripletRecipe>

Triplet recipes to use; empty means sources may provide defaults.

text_recipes: Vec<TextRecipe>

Text recipes to use; empty means derived from triplet recipes if available.

split: SplitRatios

Split ratios used when assigning records to train/val/test.

allowed_splits: Vec<SplitLabel>

Splits allowed for sampling requests.

Implementations

impl SamplerConfig

pub fn with_denoiser(self, config: DenoiserConfig) -> Self

Consuming builder to enable the built-in OCR/markdown denoiser on the sampler’s chunking strategy.

Chains denoiser setup during SamplerConfig construction. Works with struct update syntax to customize other fields at the same time:

use triplets_core::{SamplerConfig, config::DenoiserConfig};

// Enable denoiser with all other fields at their defaults:
let config = SamplerConfig::default()
    .with_denoiser(DenoiserConfig { enabled: true, ..DenoiserConfig::default() });

// Or customize other fields first, then add the denoiser:
let config = SamplerConfig { batch_size: 32, ..SamplerConfig::default() }
    .with_denoiser(DenoiserConfig { enabled: true, ..DenoiserConfig::default() });

Trait Implementations

impl Clone for SamplerConfig

fn clone(&self) -> SamplerConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SamplerConfig

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

Formats the value using the given formatter.

impl Default for SamplerConfig

fn default() -> Self

Returns the “default value” for a type.