Struct dfdx::nn::modules::Embedding

pub struct Embedding<const VOCAB: usize, const DIM: usize, E: Dtype, D: DeviceStorage> {
    pub weight: Tensor<Rank2<VOCAB, DIM>, E, D>,
}

An embedding Initializes Self::weight from a Uniform distribution between [-1 / sqrt(I), 1 / sqrt(I)].

Generics

• VOCAB The size of the vocabulary, inputs integer values must be between 0 and VOCAB;
• DIM The “output” size of vectors & matrices which are the vectors being selected.

Examples

Embedding<5, 2> can act on vectors with SEQ integer elements (with values between 0 and 4), and results in a SEQ tensor of usually f32 elements being the rows in Self::weight.

type Model = Embedding<7, 2>;
let mut model = dev.build_module::<Model, f32>();
// single sequence of ids
let inputs: Tensor<Rank1<5>, usize, _> = dev.zeros();
let _: Tensor<(Const<5>, Const<2>,), f32, _> = model.forward(inputs);
// Dynamic sequence of ids
let inputs: Tensor<(usize, ), usize, _> = dev.zeros_like(&(5, ));
let _: Tensor<(usize, Const<2>,), f32, _> = model.forward(inputs);
// batched sequence of ids
let inputs: Tensor<Rank2<10, 5>, usize, _> = dev.zeros();
let _: Tensor<(Const<10>, Const<5>, Const<2>), f32, _> = model.forward(inputs);

Fields

weight: Tensor<Rank2<VOCAB, DIM>, E, D>

Transposed weight matrix, shape (I, O)

Trait Implementations

impl<const VOCAB: usize, const DIM: usize, E: Clone + Dtype, D: Clone + DeviceStorage> Clone for Embedding<VOCAB, DIM, E, D>

fn clone(&self) -> Embedding<VOCAB, DIM, E, D>

Returns a copy of the value.

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

Performs copy-assignment from source.

impl<const VOCAB: usize, const DIM: usize, E: Debug + Dtype, D: Debug + DeviceStorage> Debug for Embedding<VOCAB, DIM, E, D>

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

Formats the value using the given formatter.

impl<const VOCAB: usize, const DIM: usize, BATCH: Dim, SEQ: Dim, E: Dtype, D: Device<E>, T: Tape<E, D>> Module<Tensor<(BATCH, SEQ), usize, D, T>> for Embedding<VOCAB, DIM, E, D>

type Output = Tensor<(BATCH, SEQ, Const<DIM>), E, D, T>

The type that this unit produces given Input.

type Error = <D as HasErr>::Err

fn try_forward( &self, input: Tensor<(BATCH, SEQ), usize, D, T> ) -> Result<Self::Output, D::Err>

fn forward(&self, input: Input) -> Self::Output

Forward Input through the module and produce Module::Output.

impl<const V: usize, const M: usize, SEQ: Dim, E: Dtype, D: Device<E>, T: Tape<E, D>> Module<Tensor<(SEQ,), usize, D, T>> for Embedding<V, M, E, D>

type Output = Tensor<(SEQ, Const<M>), E, D, T>

The type that this unit produces given Input.

type Error = <D as HasErr>::Err

fn try_forward( &self, input: Tensor<(SEQ,), usize, D, T> ) -> Result<Self::Output, D::Err>

fn forward(&self, input: Input) -> Self::Output

Forward Input through the module and produce Module::Output.

impl<const C: usize, const M: usize, E: Dtype + Float + SampleUniform, D: Device<E>> TensorCollection<E, D> for Embedding<C, M, E, D>

type To<E2: Dtype, D2: Device<E2>> = Embedding<C, M, E2, D2>

Type alias that specifies the how a module’s type changes when using a different dtype and/or device.

fn iter_tensors<V: ModuleVisitor<Self, E, D>>( visitor: &mut V ) -> Result<Option<Self::To<V::E2, V::D2>>, V::Err>

Specifies how to iterate through tensors or modules containted within this module, and how to contruct this module given values for its fields. Returns Err(_) to indicate an error, Ok(None) to indicate that there is no error and a module has not been built, and Ok(Some(_)) contains Self::Output<E2, D2>

fn module<F1, F2, Field>( name: &str, get_ref: F1, get_mut: F2 ) -> ModuleField<'_, F1, F2, Self, Field>where F1: FnMut(&Self) -> &Field, F2: FnMut(&mut Self) -> &mut Field, Field: TensorCollection<E, D>,

Creates a ModuleFields that represents a field that may contain one or more tensors.

fn tensor<F1, F2, S>( name: &str, get_ref: F1, get_mut: F2, options: TensorOptions<S, E, D> ) -> TensorField<'_, F1, F2, Self, S, E, D>where F1: FnMut(&Self) -> &Tensor<S, E, D>, F2: FnMut(&mut Self) -> &mut Tensor<S, E, D>, S: Shape,

Creates a ModuleFields that represents a tensor field.

impl<const V: usize, const M: usize, E: Dtype, D: DeviceStorage> NonMutableModule for Embedding<V, M, E, D>