Expand description
A collection of useful data utilities such as ExactSizeDataset, OneHotEncode, Arange, and iterator extension traits!
Traits
- Generates a tensor with ordered data from 0 to
N
. - Collates
Self
into some other type. Generally similar to an unzip method; Transforms[(A, B); N]
into([A; N], [B; N])
; TransformsVec<(A, B)>
into(Vec<A>, Vec<B>)
. - A dataset with a known size that can be iterated over in order or in shuffled order.
- Create batches of items from an Iterator
- One hot encodes an array of class labels into a 2d tensor of probability vectors. This can be used in tandem with crate::losses::cross_entropy_with_logits_loss().