Struct dfdx::nn::modules::Transformer

pub struct Transformer<const MODEL_DIM: usize, const NUM_HEADS: usize, const NUM_ENCODER_LAYERS: usize, const NUM_DECODER_LAYERS: usize, const FF_DIM: usize, E: Dtype, D: DeviceStorage> {
    pub encoder: TransformerEncoder<MODEL_DIM, NUM_HEADS, FF_DIM, NUM_ENCODER_LAYERS, E, D>,
    pub decoder: TransformerDecoder<MODEL_DIM, NUM_HEADS, FF_DIM, NUM_DECODER_LAYERS, E, D>,
}

Transformer architecture as described in Attention is all you need.

This is comprised of a TransformerEncoder and a TransformerDecoder.

Generics:

• MODEL_DIM: Size of the input features to the encoder/decoder.
• NUM_HEADS: Number of heads for MultiHeadAttention.
• NUM_ENCODER_LAYERS: Number of TransformerEncoderBlock to use
• NUM_DECODER_LAYERS: Number of TransformerDecoderBlock to use
• FF_DIM: Feedforward hidden dimension for both encoder/decoder

Pytorch equivalent:

torch.nn.Transformer(
    d_model=MODEL_DIM,
    nhead=NUM_HEADS,
    num_encoder_layers=NUM_ENCODER_LAYERS,
    num_decoder_layers=NUM_DECODER_LAYERS,
    dim_feedforward=FF_DIM,
    batch_first=True,
)

Fields

encoder: TransformerEncoder<MODEL_DIM, NUM_HEADS, FF_DIM, NUM_ENCODER_LAYERS, E, D>

decoder: TransformerDecoder<MODEL_DIM, NUM_HEADS, FF_DIM, NUM_DECODER_LAYERS, E, D>

Trait Implementations

impl<const MODEL_DIM: usize, const NUM_HEADS: usize, const NUM_ENCODER_LAYERS: usize, const NUM_DECODER_LAYERS: usize, const FF_DIM: usize, E: Clone + Dtype, D: Clone + DeviceStorage> Clone for Transformer<MODEL_DIM, NUM_HEADS, NUM_ENCODER_LAYERS, NUM_DECODER_LAYERS, FF_DIM, E, D>

fn clone( &self ) -> Transformer<MODEL_DIM, NUM_HEADS, NUM_ENCODER_LAYERS, NUM_DECODER_LAYERS, FF_DIM, E, D>

Returns a copy of the value.

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

Performs copy-assignment from source.

impl<const MODEL_DIM: usize, const NUM_HEADS: usize, const NUM_ENCODER_LAYERS: usize, const NUM_DECODER_LAYERS: usize, const FF_DIM: usize, E: Debug + Dtype, D: Debug + DeviceStorage> Debug for Transformer<MODEL_DIM, NUM_HEADS, NUM_ENCODER_LAYERS, NUM_DECODER_LAYERS, FF_DIM, E, D>

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

Formats the value using the given formatter.

impl<const M: usize, const H: usize, const EL: usize, const DL: usize, const F: usize, E: Dtype, D: Device<E>, Src: SplitTape, Tgt: PutTape<Src::Tape>> Module<(Src, Tgt)> for Transformer<M, H, EL, DL, F, E, D>where TransformerEncoder<M, H, F, EL, E, D>: Module<Src, Output = Src, Error = D::Err>, TransformerDecoder<M, H, F, DL, E, D>: Module<(<Tgt as PutTape<Src::Tape>>::Output, Src::NoTape), Output = <Tgt as PutTape<Src::Tape>>::Output, Error = D::Err>,

type Output = <Tgt as PutTape<<Src as SplitTape>::Tape>>::Output

The type that this unit produces given Input.

type Error = <D as HasErr>::Err

fn try_forward(&self, (src, tgt): (Src, Tgt)) -> Result<Self::Output, D::Err>

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

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

impl<const M: usize, const H: usize, const A: usize, const B: usize, const F: usize, E, D> TensorCollection<E, D> for Transformer<M, H, A, B, F, E, D>where E: Dtype + Float + SampleUniform, D: Device<E>,

type To<E2: Dtype, D2: Device<E2>> = Transformer<M, H, A, B, F, 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 M: usize, const H: usize, const A: usize, const B: usize, const F: usize, E, D> NonMutableModule for Transformer<M, H, A, B, F, E, D>where E: Dtype, D: Device<E>,