Trait Dataset

pub trait Dataset {
    type Item;

    // Required methods
    fn len(&self) -> usize;
    fn get(&self, index: usize) -> Self::Item;

    // Provided methods
    fn cached(self, cache_dir: impl Into<PathBuf>) -> CachedDataset<Self>
       where Self: Sized { ... }
    fn iter(&self) -> Box<dyn Iterator<Item = Self::Item> + '_> { ... }
}

Generic Dataset trait, modeled after PyTorch’s utils.data.Dataset. It represents a collection of objects indexed in the range 0..len().

Required Associated Types

type Item

The type of objects procured by the Dataset.

Required Methods

fn len(&self) -> usize

The number of objects in the Dataset.

fn get(&self, index: usize) -> Self::Item

Get the index’th element, where 0 <= index < len()

Provided Methods

fn cached(self, cache_dir: impl Into<PathBuf>) -> CachedDataset<Self>

where Self: Sized,

Modifies the dataset to check a cache directory before reading. If the cache entry is present, it’s used instead of the underlying get(). If the cache entry is absent, it will be created after calling get().

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

Produces an Iterator that produces the entire Dataset in indexed order.

Implementations on Foreign Types

impl<D: Dataset + ?Sized> Dataset for &D

References to Datasets are also Datasets.

type Item = <D as Dataset>::Item

fn len(&self) -> usize

fn get(&self, index: usize) -> Self::Item

impl<D: Dataset + ?Sized> Dataset for Box<D>

type Item = <D as Dataset>::Item

fn len(&self) -> usize

fn get(&self, index: usize) -> Self::Item

impl<T: Clone> Dataset for [T]

A slice can act as an in-memory Dataset.

type Item = T

fn len(&self) -> usize

fn get(&self, index: usize) -> T

Implementors

impl<D: Dataset> Dataset for CachedDataset<D>

where D::Item: Serialize + DeserializeOwned,

type Item = <D as Dataset>::Item

impl<T, F: Fn(usize) -> T> Dataset for ClosureDataset<T, F>

type Item = T