com.intel.analytics.bigdl.nn.tf.BiasAdd.scala Maven / Gradle / Ivy

/*
 * Copyright 2016 The BigDL Authors.
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */
package com.intel.analytics.bigdl.nn.tf

import com.intel.analytics.bigdl.nn.abstractnn.AbstractModule
import com.intel.analytics.bigdl.tensor.TensorNumericMath.TensorNumeric
import com.intel.analytics.bigdl.tensor._
import com.intel.analytics.bigdl.utils.Table

import scala.reflect.ClassTag

class BiasAdd[T: ClassTag]()
  (implicit ev: TensorNumeric[T]) extends AbstractModule[Table, Tensor[T], T] {
  var onesBias: Tensor[T] = _

  override def updateOutput(input: Table): Tensor[T] = {
    val value = input[Tensor[T]](1)
    val bias = input[Tensor[T]](2)
    val sizes = value.size().toBuffer
    val last = sizes.last
    sizes.remove(value.nDimension() - 1)
    val sizeProduct = sizes.product

    if (value.getType() != output.getType()) {
      output = value.emptyInstance()
    }

    if (onesBias == null) {
      onesBias = value.emptyInstance()
    }

    if (onesBias.dim() != 1 || onesBias.size(1) != sizeProduct) {
      onesBias.resize(sizeProduct).fill(ev.fromType(1.0))
    }

    output.resizeAs(value)
      .copy(value)
    val value2d = output.view(Array(sizeProduct, last))


    value2d
      .addr(
        value.getTensorNumeric().one,
        onesBias,
        bias)
    output
  }

  override def updateGradInput(input: Table, gradOutput: Tensor[T]): Table = {
    val value = input[Tensor[T]](1)
    val bias = input[Tensor[T]](2)

    val sizes = value.size().toBuffer
    val last = sizes.last
    sizes.remove(value.nDimension() - 1)
    val sizeProduct = sizes.product

    if (!gradInput.contains(1)) {
      gradInput(1) = value.emptyInstance()
    }

    if (!gradInput.contains(2)) {
      gradInput(2) = bias.emptyInstance()
    }

    val gradValue = gradInput[Tensor[T]](1)
    val gradBias = gradInput[Tensor[T]](2)

    gradValue.resizeAs(value).copy(gradOutput)

    val gradOutput2d = gradOutput.view(Array(sizeProduct, last))

    gradBias.resizeAs(bias).addmv(ev.fromType(1.0), gradOutput2d.t, onesBias)

    gradInput
  }
}

object BiasAdd {
  def apply[T: ClassTag]()
    (implicit ev: TensorNumeric[T]):
  BiasAdd[T]
  = new BiasAdd[T]()
}