org.deeplearning4j.nn.layers.convolution

Class CudnnConvolutionHelper

java.lang.Object

org.deeplearning4j.nn.layers.BaseCudnnHelper

org.deeplearning4j.nn.layers.convolution.CudnnConvolutionHelper

All Implemented Interfaces:

org.deeplearning4j.nn.layers.convolution.ConvolutionHelper, org.deeplearning4j.nn.layers.LayerHelper

public class CudnnConvolutionHelper
extends BaseCudnnHelper
implements org.deeplearning4j.nn.layers.convolution.ConvolutionHelper

cuDNN-based helper for the convolution layer.

Author:

saudet

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	CudnnConvolutionHelper.CudnnForwardArgs

Nested classes/interfaces inherited from class org.deeplearning4j.nn.layers.BaseCudnnHelper

BaseCudnnHelper.CudnnContext, BaseCudnnHelper.DataCache, BaseCudnnHelper.TensorArray

Field Summary

Fields inherited from class org.deeplearning4j.nn.layers.BaseCudnnHelper

alpha, beta, dataType, dataTypeSize, sizeInBytes, TENSOR_FORMAT

Constructor Summary

Constructors

Constructor and Description
CudnnConvolutionHelper()

Method Summary

Modifier and Type	Method and Description
INDArray	activate(INDArray z, IActivation afn)
org.nd4j.linalg.primitives.Pair<org.deeplearning4j.nn.gradient.Gradient,INDArray>	backpropGradient(INDArray input, INDArray weights, INDArray delta, int[] kernel, int[] strides, int[] pad, INDArray biasGradView, INDArray weightGradView, IActivation afn, org.deeplearning4j.nn.conf.layers.ConvolutionLayer.AlgoMode mode, org.deeplearning4j.nn.conf.layers.ConvolutionLayer.BwdFilterAlgo bwdFilterAlgo, org.deeplearning4j.nn.conf.layers.ConvolutionLayer.BwdDataAlgo bwdDataAlgo, org.deeplearning4j.nn.conf.ConvolutionMode convolutionMode, int[] dilation, org.deeplearning4j.nn.workspace.LayerWorkspaceMgr workspaceMgr)
static CudnnConvolutionHelper.CudnnForwardArgs	getCudnnForwardArgs(INDArray input, int[] kernel, int[] strides, int[] padding, int[] dilation, org.deeplearning4j.nn.conf.ConvolutionMode convolutionMode, org.deeplearning4j.nn.conf.layers.PoolingType poolingType)
Map<String,Long>	helperMemoryUse()
INDArray	preOutput(INDArray input, INDArray weights, INDArray bias, int[] kernel, int[] strides, int[] pad, org.deeplearning4j.nn.conf.layers.ConvolutionLayer.AlgoMode mode, org.deeplearning4j.nn.conf.layers.ConvolutionLayer.FwdAlgo fwdAlgo, org.deeplearning4j.nn.conf.ConvolutionMode convolutionMode, int[] dilation, org.deeplearning4j.nn.workspace.LayerWorkspaceMgr workspaceMgr)

Methods inherited from class org.deeplearning4j.nn.layers.BaseCudnnHelper

adaptForTensorDescr, checkCuda, checkCudnn, checkSupported, toCudnnDataType

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface org.deeplearning4j.nn.layers.convolution.ConvolutionHelper

checkSupported

Constructor Detail

CudnnConvolutionHelper

public CudnnConvolutionHelper()

Method Detail

backpropGradient

public org.nd4j.linalg.primitives.Pair<org.deeplearning4j.nn.gradient.Gradient,INDArray> backpropGradient(INDArray input,
                                                                                                          INDArray weights,
                                                                                                          INDArray delta,
                                                                                                          int[] kernel,
                                                                                                          int[] strides,
                                                                                                          int[] pad,
                                                                                                          INDArray biasGradView,
                                                                                                          INDArray weightGradView,
                                                                                                          IActivation afn,
                                                                                                          org.deeplearning4j.nn.conf.layers.ConvolutionLayer.AlgoMode mode,
                                                                                                          org.deeplearning4j.nn.conf.layers.ConvolutionLayer.BwdFilterAlgo bwdFilterAlgo,
                                                                                                          org.deeplearning4j.nn.conf.layers.ConvolutionLayer.BwdDataAlgo bwdDataAlgo,
                                                                                                          org.deeplearning4j.nn.conf.ConvolutionMode convolutionMode,
                                                                                                          int[] dilation,
                                                                                                          org.deeplearning4j.nn.workspace.LayerWorkspaceMgr workspaceMgr)

Specified by:

backpropGradient in interface org.deeplearning4j.nn.layers.convolution.ConvolutionHelper

preOutput

public INDArray preOutput(INDArray input,
                          INDArray weights,
                          INDArray bias,
                          int[] kernel,
                          int[] strides,
                          int[] pad,
                          org.deeplearning4j.nn.conf.layers.ConvolutionLayer.AlgoMode mode,
                          org.deeplearning4j.nn.conf.layers.ConvolutionLayer.FwdAlgo fwdAlgo,
                          org.deeplearning4j.nn.conf.ConvolutionMode convolutionMode,
                          int[] dilation,
                          org.deeplearning4j.nn.workspace.LayerWorkspaceMgr workspaceMgr)

Specified by:

preOutput in interface org.deeplearning4j.nn.layers.convolution.ConvolutionHelper

activate

public INDArray activate(INDArray z,
                         IActivation afn)

Specified by:

activate in interface org.deeplearning4j.nn.layers.convolution.ConvolutionHelper

getCudnnForwardArgs

public static CudnnConvolutionHelper.CudnnForwardArgs getCudnnForwardArgs(INDArray input,
                                                                          int[] kernel,
                                                                          int[] strides,
                                                                          int[] padding,
                                                                          int[] dilation,
                                                                          org.deeplearning4j.nn.conf.ConvolutionMode convolutionMode,
                                                                          org.deeplearning4j.nn.conf.layers.PoolingType poolingType)

Parameters:

poolingType - Used when preparing data for subsampling layers ONLY. Null for convolution layers

Returns:

helperMemoryUse

public Map<String,Long> helperMemoryUse()

Specified by:

helperMemoryUse in interface org.deeplearning4j.nn.layers.LayerHelper