Struct InstructDataset 

pub struct InstructDataset { /* private fields */ }

Instruction-tuning dataset (multi-turn conversations).

Implementations

impl InstructDataset

pub fn new(examples: Vec<TrainingExample>) -> Self

Create a new instruct dataset from a vector of examples.

pub fn from_examples( examples: impl IntoIterator<Item = TrainingExample>, ) -> Self

Create a new instruct dataset from an iterator of examples.

pub fn push(&mut self, example: TrainingExample)

Append a single example to the dataset.

pub fn extend(&mut self, examples: impl IntoIterator<Item = TrainingExample>)

Extend the dataset with an iterator of examples.

pub fn remove(&mut self, index: usize) -> DatasetResult<TrainingExample>

Remove and return the example at the given index.

pub fn total_estimated_tokens(&self) -> usize

Total estimated tokens across all examples.

pub fn filter<F>(&self, predicate: F) -> Self

where F: Fn(&TrainingExample) -> bool,

Filter examples by a predicate.

pub fn as_slice(&self) -> &[TrainingExample]

Get all examples as a slice.

pub fn into_inner(self) -> Vec<TrainingExample>

Consume self and return the underlying Vec.

Trait Implementations

impl Clone for InstructDataset

fn clone(&self) -> InstructDataset

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Dataset for InstructDataset

type Item = TrainingExample

The item type stored in this dataset.

fn len(&self) -> usize

Return the number of items in the dataset.

fn get(&self, index: usize) -> Option<&TrainingExample>

Get an item by index.

fn iter(&self) -> Box<dyn Iterator<Item = &TrainingExample> + '_>

Return an iterator over all items.

fn shuffle(&mut self, seed: u64)

Shuffle the dataset in place using the given seed.

fn split(&self, ratio: f32) -> (Vec<TrainingExample>, Vec<TrainingExample>)

Split the dataset by ratio into two vectors.

fn is_empty(&self) -> bool

Return true if the dataset is empty.

impl Debug for InstructDataset

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

Formats the value using the given formatter.