org.nd4j.linalg.factory.Nd4j.toFlattened()方法的使用及代码示例

x33g5p2x  于2022-01-24 转载在 其他  
字(7.6k)|赞(0)|评价(0)|浏览(166)

本文整理了Java中org.nd4j.linalg.factory.Nd4j.toFlattened()方法的一些代码示例，展示了Nd4j.toFlattened()的具体用法。这些代码示例主要来源于Github/Stackoverflow/Maven等平台，是从一些精选项目中提取出来的代码，具有较强的参考意义，能在一定程度帮忙到你。Nd4j.toFlattened()方法的具体详情如下：
包路径：org.nd4j.linalg.factory.Nd4j
类名称：Nd4j
方法名：toFlattened

Nd4j.toFlattened介绍

[英]Create a long row vector of all of the given ndarrays
[中]创建所有给定nArray的长行向量

代码示例

代码示例来源：origin: deeplearning4j/dl4j-examples

  1. static INDArray append(INDArray arr1, INDArray values, int dimension) {
  2.   if(dimension == -1) {
  3.     return Nd4j.toFlattened(arr1, values);
  4.   } else {
  5.     return Nd4j.concat(dimension, arr1, values);
  6.   }
  7. }

代码示例来源：origin: deeplearning4j/dl4j-examples

  1. static INDArray insert(INDArray arr1, int index, INDArray values, int dimension) {
  2.   if(dimension == -1) {
  3.     INDArray flat1 = Nd4j.toFlattened(arr1);
  4.     INDArray flatValues = Nd4j.toFlattened(values);
  5.     INDArray firstSlice = flat1.get(NDArrayIndex.interval(0, index));
  6.     INDArray secondSlice = flat1.get(NDArrayIndex.interval(index, flat1.length()));
  7.     return Nd4j.toFlattened(firstSlice, flatValues, secondSlice);
  8.   } else {
  9.     INDArray firstSlice = arr1.get(createIntervalOnDimension(dimension, false,
  10.       0, index));
  11.     INDArray secondSlice = arr1.get(createIntervalOnDimension(dimension, false,
  12.       index, arr1.shape()[dimension]));
  13.     return Nd4j.concat(dimension, firstSlice, values, secondSlice);
  14.   }
  15. }

代码示例来源：origin: deeplearning4j/nd4j

  1. protected INDArray handleParamsView(INDArray outputArray, INDArray paramView) {
  2.   //minor optimization when the views are the same, just return
  3.   if(paramView == null || paramView == outputArray)
  4.     return outputArray;
  5.   INDArray flat = Nd4j.toFlattened(order(), outputArray);
  6.   if (flat.length() != paramView.length())
  7.     throw new RuntimeException("ParamView length does not match initialized weights length (view length: "
  8.         + paramView.length() + ", view shape: " + Arrays.toString(paramView.shape())
  9.         + "; flattened length: " + flat.length());
  10.   paramView.assign(flat);
  11.   return paramView.reshape(order(), outputArray.shape());
  12. }

代码示例来源：origin: deeplearning4j/nd4j

  1. public static INDArray tailor4d2d(@NonNull INDArray data) {
  2.   long instances = data.size(0);
  3.   long channels = data.size(1);
  4.   long height = data.size(2);
  5.   long width = data.size(3);
  6.   INDArray in2d = Nd4j.create(channels, height * width * instances);
  7.   long tads = data.tensorssAlongDimension(3, 2, 0);
  8.   for (int i = 0; i < tads; i++) {
  9.     INDArray thisTAD = data.tensorAlongDimension(i, 3, 2, 0);
  10.     in2d.putRow(i, Nd4j.toFlattened(thisTAD));
  11.   }
  12.   return in2d.transposei();
  13. }

代码示例来源：origin: deeplearning4j/dl4j-examples

  1. static INDArray delete(int dimension, INDArray arr1, int... interval) {
  2.   int length = interval.length;
  3.   int lastIntervalValue = interval[length - 1];
  4.   if(dimension == -1) {
  5.     INDArray array1 = arr1.get(NDArrayIndex.interval(0, interval[0]));
  6.     if(lastIntervalValue == arr1.length() - 1) {
  7.       return Nd4j.toFlattened(array1);
  8.     } else {
  9.       INDArray array2 = arr1.get(NDArrayIndex.interval(lastIntervalValue + 1,
  10.         arr1.length()));
  11.       return Nd4j.toFlattened(array1, array2);
  12.     }
  13.   } else {
  14.     INDArray array1 = arr1.get(createIntervalOnDimension(dimension, false, 0, interval[0]));
  15.     if(lastIntervalValue == arr1.shape()[dimension] - 1) {
  16.       return array1;
  17.     } else {
  18.       INDArray array2 = arr1.get(createIntervalOnDimension(dimension, false,
  19.         lastIntervalValue + 1,
  20.         arr1.shape()[dimension]));
  21.       return Nd4j.concat(dimension, array1, array2);
  22.     }
  23.   }
  24. }

代码示例来源：origin: deeplearning4j/dl4j-examples

  1. print("Ascended sorted array on zero axis: ", axisSortedArray);
  2. INDArray flattened = Nd4j.toFlattened(fourByFiveRandomZeroToOne);
  3. print("Flattened array", flattened);

代码示例来源：origin: org.deeplearning4j/deeplearning4j-nn

  1. @Override
  2. public INDArray params() {
  3.   //C order flattening, to match the gradient flattening order
  4.   return Nd4j.toFlattened('c', params.values());
  5. }

代码示例来源：origin: org.deeplearning4j/deeplearning4j-nn

  1. public INDArray params() {
  2.   List<INDArray> list = new ArrayList<>(2);
  3.   for (Map.Entry<String, INDArray> entry : params.entrySet()) {
  4.     list.add(entry.getValue());
  5.   }
  6.   return Nd4j.toFlattened('f', list);
  7. }

代码示例来源：origin: org.deeplearning4j/deeplearning4j-nn

  1. private void flattenGradient() {
  2.   if (flatteningOrders != null) {
  3.     //Arrays with non-default order get flattened to row vector first, then everything is flattened to f order
  4.     //TODO revisit this, and make more efficient
  5.     List<INDArray> toFlatten = new ArrayList<>();
  6.     for (Map.Entry<String, INDArray> entry : gradients.entrySet()) {
  7.       if (flatteningOrders.containsKey(entry.getKey())
  8.               && flatteningOrders.get(entry.getKey()) != DEFAULT_FLATTENING_ORDER) {
  9.         //Specific flattening order for this array, that isn't the default
  10.         toFlatten.add(Nd4j.toFlattened(flatteningOrders.get(entry.getKey()), entry.getValue()));
  11.       } else {
  12.         //default flattening order for this array
  13.         toFlatten.add(entry.getValue());
  14.       }
  15.     }
  16.     flattenedGradient = Nd4j.toFlattened(DEFAULT_FLATTENING_ORDER, toFlatten);
  17.   } else {
  18.     //Standard case: flatten all to f order
  19.     flattenedGradient = Nd4j.toFlattened(DEFAULT_FLATTENING_ORDER, gradients.values());
  20.   }
  21. }

代码示例来源：origin: org.deeplearning4j/deeplearning4j-nn

  1. @Override
  2. public INDArray gradient(List<String> order) {
  3.   List<INDArray> toFlatten = new ArrayList<>();
  4.   if (flatteningOrders == null) {
  5.     for (String s : order) {
  6.       if (!gradients.containsKey(s))
  7.         continue;
  8.       toFlatten.add(gradients.get(s));
  9.     }
  10.   } else {
  11.     for (String s : order) {
  12.       if (!gradients.containsKey(s))
  13.         continue;
  14.       if (flatteningOrders.containsKey(s) && flatteningOrders.get(s) != DEFAULT_FLATTENING_ORDER) {
  15.         //Arrays with non-default order get flattened to row vector first, then everything is flattened to f order
  16.         //TODO revisit this, and make more efficient
  17.         toFlatten.add(Nd4j.toFlattened(flatteningOrders.get(s), gradients.get(s)));
  18.       } else {
  19.         toFlatten.add(gradients.get(s));
  20.       }
  21.     }
  22.   }
  23.   return Nd4j.toFlattened(DEFAULT_FLATTENING_ORDER, toFlatten);
  24. }

代码示例来源：origin: org.deeplearning4j/deeplearning4j-nn

  1. public INDArray scoreExamples(DataSetIterator iter, boolean addRegularizationTerms) {
  2.   List<INDArray> out = new ArrayList<>();
  3.   while (iter.hasNext()) {
  4.     out.add(scoreExamples(iter.next(), addRegularizationTerms));
  5.   }
  6.   return Nd4j.toFlattened('f', out);
  7. }

代码示例来源：origin: org.deeplearning4j/deeplearning4j-nn

  1. /**
  2.  * Returns a 1 x m vector where the vector is composed of
  3.  * a flattened vector of all of the weights for the
  4.  * various neuralNets(w,hbias NOT VBIAS) and output layer
  5.  *
  6.  * @return the params for this neural net
  7.  */
  8. public INDArray params(boolean backwardOnly) {
  9.   if (backwardOnly)
  10.     return params();
  11.   List<INDArray> params = new ArrayList<>();
  12.   for (Layer layer : getLayers()) {
  13.     INDArray layerParams = layer.params();
  14.     if (layerParams != null)
  15.       params.add(layerParams); //may be null: subsampling etc layers
  16.   }
  17.   return Nd4j.toFlattened('f', params);
  18. }

代码示例来源：origin: org.deeplearning4j/deeplearning4j-nn

  1. /**
  2.  * Get the parameters for the ComputationGraph
  3.  *
  4.  * @param backwardOnly If true: backprop parameters only (i.e., no visible layer biases used in layerwise pretraining layers)
  5.  */
  6. public INDArray params(boolean backwardOnly) {
  7.   if (backwardOnly)
  8.     return flattenedParams;
  9.   List<INDArray> list = new ArrayList<>(layers.length);
  10.   for (int i = 0; i < topologicalOrder.length; i++) {
  11.     if (!vertices[topologicalOrder[i]].hasLayer())
  12.       continue;
  13.     Layer l = vertices[topologicalOrder[i]].getLayer();
  14.     INDArray layerParams = l.params();
  15.     if (layerParams != null)
  16.       list.add(layerParams); //may be null: subsampling etc layers
  17.   }
  18.   return Nd4j.toFlattened('f', list);
  19. }

代码示例来源：origin: org.nd4j/nd4j-api

  1. public static INDArray tailor4d2d(@NonNull INDArray data) {
  2.   int instances = data.size(0);
  3.   int channels = data.size(1);
  4.   int height = data.size(2);
  5.   int width = data.size(3);
  6.   INDArray in2d = Nd4j.create(channels, height * width * instances);
  7.   int tads = data.tensorssAlongDimension(3, 2, 0);
  8.   for (int i = 0; i < tads; i++) {
  9.     INDArray thisTAD = data.tensorAlongDimension(i, 3, 2, 0);
  10.     in2d.putRow(i, Nd4j.toFlattened(thisTAD));
  11.   }
  12.   return in2d.transposei();
  13. }

代码示例来源：origin: Waikato/wekaDeeplearning4j

  1. for (int i = 0; i < batchsize; i++) {
  2.  INDArray row = ndArray.getRow(i);
  3.  INDArray flattenedRow = Nd4j.toFlattened(row);
  4.  Instance inst = new DenseInstance(atts.size());
  5.  for (int j = 0; j < flattenedRow.size(1); j++) {

代码示例来源：origin: org.deeplearning4j/deeplearning4j-nn

  1. INDArray flat = Nd4j.toFlattened(order, ret);
  2. if (flat.length() != paramView.length())
  3.   throw new RuntimeException("ParamView length does not match initialized weights length (view length: "

内容来源于网络，如有侵权，请联系作者删除！

Nd4j

相关文章

    查看更多

    热门标签

    更多
    JavaquerypythonNode开发语言requestUtil数据库Table后端算法LoggerMessageElementParser

    最新文章

    更多

    Nd4j类方法

    • 蜀ICP备13028337号-1 大数据知识库 https://www.saoniuhuo.com © All rights reserved
    • 本站内容来源互联网,如果侵犯您的权益请联系我们删除, 联系方式：448109455@qq.com