nlpodyssey/spago

// Copyright 2019 spaGO Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

// Package layernorm implements the Layer Normalization (LayerNorm) i method.
//
// Reference: "Layer normalization" by Jimmy Lei Ba, Jamie Ryan Kiros, and Geoffrey E Hinton (2016).
// (https://arxiv.org/pdf/1607.06450.pdf)
package layernorm

import (
    "encoding/gob"

    "github.com/nlpodyssey/spago/ag"
    "github.com/nlpodyssey/spago/mat"
    "github.com/nlpodyssey/spago/mat/float"
    "github.com/nlpodyssey/spago/nn"
)

var _ nn.Model = &Model{}

// Model contains the serializable parameters.
type Model struct {
    nn.Module
    W   *nn.Param
    B   *nn.Param
    Eps *nn.Buffer
}

func init() {
    gob.Register(&Model{})
}

// New returns a new model with parameters initialized to zeros.
func New[T float.DType](size int, eps float64) *Model {
    return &Model{
        W:   nn.NewParam(mat.NewDense[T](mat.WithShape(size))),
        B:   nn.NewParam(mat.NewDense[T](mat.WithShape(size))),
        Eps: nn.Buf(mat.Scalar(T(eps))),
    }
}

// Forward performs the forward step for each input node and returns the result.
// y = (x - E\[x\]) / sqrt(VAR\[x\] + [EPS]) * g + b
func (m *Model) Forward(xs ...mat.Tensor) []mat.Tensor {
    if len(xs) == 0 {
        return nil
    }
    out := make([]mat.Tensor, len(xs))
    for i, x := range xs {
        mean := ag.ReduceMean(x)
        dev := ag.SubScalar(x, mean)
        stdDev := ag.Sqrt(ag.Add(ag.ReduceMean(ag.Square(dev)), m.Eps))
        out[i] = ag.Add(ag.Prod(ag.DivScalar(dev, stdDev), m.W), m.B)
    }
    return out
}