caffe源码导读(五)Layer数据结构描述

本篇为基于<深度学习21天实战caffe>所做的学习笔记,如果错误,欢迎指导

一、前篇

caffe 源码导读(一) 了解protobuf
caffe 源码导读(二) Blob数据结构介绍
caffe 源码导读(三) Blob.hpp头文件解析
caffe 源码导读(四) Blob.cpp解析
d===========================================================================d

二、正文

Layer是caffe的基本计算单元,至少有一个输入Blob和输出Blob,部分Layer带有权值和偏置. 有两个运算方向:前向传播和反向传播,其中前向传播计算会对输入Blob进行某种处理(有权值和偏置项的Layer会利用这些对输入进行处理),得到输出Blob,而反向传播计算则对输出Blob的diff进行某种处理,得到输入Blob的diff(有权值和偏置项的Layer可能也会计算权值Blob\偏置项Blob的diff).

2.1 LayerParameter的数据结构描述(Proto)

看Proto代码,发现还包含Phase、ParamSpec等

// NOTE
// Update the next available ID when you add a new LayerParameter field.
// 注意:当你怎增加一个新的Layerparameter 字段时,更新一下下一个可用的ID
// LayerParameter next available layer-specific ID: 149 (last added: clip_param)
message LayerParameter {
  optional string name = 1; // the layer name Layer名称
  optional string type = 2; // the layer type Layer类型
  repeated string bottom = 3; // the name of each bottom blob输入(Bootom Blob)的名称
  repeated string top = 4; // the name of each top blob blob输出(Top Blob)的名称

  // The train / test phase for computation.
  optional Phase phase = 10;  // 阶段 (TRAIN/TEST)

  // The amount of weight to assign each top blob in the objective.
  // Each layer assigns a default value, usually of either 0 or 1,
  // to each top blob.
  // 为每个TopBlob 分配对损失函数的权重,每个Layer都有默认值,要么为0,表示不参与损失函数的计算,要么为1
  // 表示参与损失函数的计算
  repeated float loss_weight = 5;

  // Specifies training parameters (multipliers on global learning constants,
  // and the name and other settings used for weight sharing).
  // 指定的训练参数,例如 相对全局学习常数的缩放因子,以及用于权值共享的名称或其他设置,比如lr_mult decay_mult等
  repeated ParamSpec param = 6;

  // The blobs containing the numeric parameters of the layer.
  // 承载该层参数的blob
  repeated BlobProto blobs = 7;

  // Specifies whether to backpropagate to each bottom. If unspecified,
  // Caffe will automatically infer whether each input needs backpropagation
  // to compute parameter gradients. If set to true for some inputs,
  // backpropagation to those inputs is forced; if set false for some inputs,
  // backpropagation to those inputs is skipped.
  //指定是否对Bootom Blob进行反向传播过程,该字段的维度应该与Bottom Blob个数一致
  // 如果不指定,caffe会自动检查腿短每一个输入是否都需要反向传播计算,
  // 如果一些输入设置为true,这些layer将强制反向传播
  // 如果设置为false,这些layer将跳过反向传播
  // The size must be either 0 or equal to the number of bottoms.
  repeated bool propagate_down = 11;

  // Rules controlling whether and when a layer is included in the network,
  // based on the current NetState.  You may specify a non-zero number of rules
  // to include OR exclude, but not both.  If no include or exclude rules are
  // specified, the layer is always included.  If the current NetState meets
  // ANY (i.e., one or more) of the specified rules, the layer is
  // included/excluded.
  // 控制某个层某个时刻是否包含在网络中,基于当前的NetSatateRule.
  // 可以为include或者exclude,如果没有任何规则,那个该层一致包含在网络中.其他根据指定来
  repeated NetStateRule include = 8;
  repeated NetStateRule exclude = 9;

  // Parameters for data pre-processing.
  // 数据预处理参数
  optional TransformationParameter transform_param = 100;

  // Parameters shared by loss layers.
  // 损失层共享的参数
  optional LossParameter loss_param = 101;

  // Layer type-specific parameters.
  //
  // Note: certain layers may have more than one computational engine
  // for their implementation. These layers include an Engine type and
  // engine parameter for selecting the implementation.
  // The default for the engine is set by the ENGINE switch at compile-time.
  // 特定类型层的参数.
  optional AccuracyParameter accuracy_param = 102;
  optional ArgMaxParameter argmax_param = 103;
  optional BatchNormParameter batch_norm_param = 139;
  optional BiasParameter bias_param = 141;
  optional ClipParameter clip_param = 148; //下一个是149
  optional ConcatParameter concat_param = 104;
  optional ContrastiveLossParameter contrastive_loss_param = 105;
  optional ConvolutionParameter convolution_param = 106;
  optional CropParameter crop_param = 144;
  optional DataParameter data_param = 107;
  optional DropoutParameter dropout_param = 108;
  optional DummyDataParameter dummy_data_param = 109;
  optional EltwiseParameter eltwise_param = 110;
  optional ELUParameter elu_param = 140;
  optional EmbedParameter embed_param = 137;
  optional ExpParameter exp_param = 111;
  optional FlattenParameter flatten_param = 135;
  optional HDF5DataParameter hdf5_data_param = 112;
  optional HDF5OutputParameter hdf5_output_param = 113;
  optional HingeLossParameter hinge_loss_param = 114;
  optional ImageDataParameter image_data_param = 115;
  optional InfogainLossParameter infogain_loss_param = 116;
  optional InnerProductParameter inner_product_param = 117;
  optional InputParameter input_param = 143;
  optional LogParameter log_param = 134;
  optional LRNParameter lrn_param = 118;
  optional MemoryDataParameter memory_data_param = 119;
  optional MVNParameter mvn_param = 120;
  optional ParameterParameter parameter_param = 145;
  optional PoolingParameter pooling_param = 121;
  optional PowerParameter power_param = 122;
  optional PReLUParameter prelu_param = 131;
  optional PythonParameter python_param = 130;
  optional RecurrentParameter recurrent_param = 146;
  optional ReductionParameter reduction_param = 136;
  optional ReLUParameter relu_param = 123;
  optional ReshapeParameter reshape_param = 133;
  optional ScaleParameter scale_param = 142;
  optional SigmoidParameter sigmoid_param = 124;
  optional SoftmaxParameter softmax_param = 125;
  optional SPPParameter spp_param = 132;
  optional SliceParameter slice_param = 126;
  optional SwishParameter swish_param = 147;
  optional TanHParameter tanh_param = 127;
  optional ThresholdParameter threshold_param = 128;
  optional TileParameter tile_param = 138;
  optional WindowDataParameter window_data_param = 129;
}

2.2 Phase结构

就两个 TRAIN和 TEST

enum Phase {
   TRAIN = 0;
   TEST = 1;
}

2.3 ParamSpec结构

// Specifies training parameters (multipliers on global learning constants,
// and the name and other settings used for weight sharing).
// 指定训练参数以及权重共享的其他设置
message ParamSpec {
  // The names of the parameter blobs -- useful for sharing parameters among
  // layers, but never required otherwise.  To share a parameter between two
  // layers, give it a (non-empty) name.
  // 两个参数之间进行参数共享的blob的名字
  optional string name = 1;

  // Whether to require shared weights to have the same shape, or just the same
  // count -- defaults to STRICT if unspecified.
  // 参数共享时是否需要相同的shape,默认是需要相同的
  optional DimCheckMode share_mode = 2;
  enum DimCheckMode {
    // STRICT (default) requires that num, channels, height, width each match.
    STRICT = 0;
    // PERMISSIVE requires only the count (num*channels*height*width) to match.
    PERMISSIVE = 1;
  }

  // The multiplier on the global learning rate for this parameter.
  // 学习率参数,learning rate = base+lr * lr_mult
  optional float lr_mult = 3 [default = 1.0];

  // The multiplier on the global weight decay for this parameter.
  // 权重衰减系数, weight = weight_decay * dacay_mult
  optional float decay_mult = 4 [default = 1.0];
}

2.4 NetStateRule结构

// 网络状态
message NetStateRule {
  // Set phase to require the NetState have a particular phase (TRAIN or TEST)
  // to meet this rule.
  // 设置phase
  optional Phase phase = 1;

  // Set the minimum and/or maximum levels in which the layer should be used.
  // Leave undefined to meet the rule regardless of level.
  // 设置Layer的level
  optional int32 min_level = 2;
  optional int32 max_level = 3;

  // Customizable sets of stages to include or exclude.
  // The net must have ALL of the specified stages and NONE of the specified
  // "not_stage"s to meet the rule.
  // (Use multiple NetStateRules to specify conjunctions of stages.)
  // 可定制的stage集合
  repeated string stage = 4;
  repeated string not_stage = 5;
}

2.5 TransformationParameter 结构

// Message that stores parameters used to apply transformation
// to the data layer's data
// 用来进行数据层(图像)变换的参数
message TransformationParameter {
  // For data pre-processing, we can do simple scaling and subtracting the
  // data mean, if provided. Note that the mean subtraction is always carried
  // out before scaling.
  // 像素归一化,归一化前会减去均值
  optional float scale = 1 [default = 1];
  // Specify if we want to randomly mirror data.
  // 图像随机mirror操作
  optional bool mirror = 2 [default = false];
  // Specify if we would like to randomly crop an image.
  // 图像随机crop操作
  optional uint32 crop_size = 3 [default = 0];
  // mean_file and mean_value cannot be specified at the same time
  // 图像的均值文件
  optional string mean_file = 4;
  // if specified can be repeated once (would subtract it from all the channels)
  // or can be repeated the same number of times as channels
  // (would subtract them from the corresponding channel)
  // 图像的均值,手动指定,通常三个
  repeated float mean_value = 5;
  // Force the decoded image to have 3 color channels.
  // 强制图像必须有三个通道
  optional bool force_color = 6 [default = false];
  // Force the decoded image to have 1 color channels.
  // 强制图像还有一个通道(灰度图像)
  optional bool force_gray = 7 [default = false];
}

2.6 LossParameter 结构

// Message that stores parameters shared by loss layers
// loss层的参数
message LossParameter {
  // If specified, ignore instances with the given label.
  // 如果指定,则label等于ignore_label的样本将不参与Loss计算,并且反向传播
  // 时的梯度直接置0
  optional int32 ignore_label = 1;
  // How to normalize the loss for loss layers that aggregate across batches,
  // spatial dimensions, or other dimensions.  Currently only implemented in
  // SoftmaxWithLoss and SigmoidCrossEntropyLoss layers.
  // 指定loss归一化的方式,目前只在SoftmaxWithLoss and SigmoidCrossEntropyLoss上实施
  enum NormalizationMode {
    // Divide by the number of examples in the batch times spatial dimensions.
    // Outputs that receive the ignore label will NOT be ignored in computing
    // the normalization factor.
    // 所有样本都参与计算,包括ignore label
    FULL = 0;
    // Divide by the total number of output locations that do not take the
    // ignore_label.  If ignore_label is not set, this behaves like FULL.
    // 所有样本都参与计算,不包括ignore label
    VALID = 1;
    // Divide by the batch size.
    // 除以给定的batch size
    BATCH_SIZE = 2;
    // Do not normalize the loss.
    // 不归一化 loss
    NONE = 3;
  }
  // For historical reasons, the default normalization for
  // SigmoidCrossEntropyLoss is BATCH_SIZE and *not* VALID.
  // loss归一化方式
  optional NormalizationMode normalization = 3 [default = VALID];
  // Deprecated.  Ignored if normalization is specified.  If normalization
  // is not specified, then setting this to false will be equivalent to
  // normalization = BATCH_SIZE to be consistent with previous behavior.
  // 已经放弃,loss会除以参与计算样本的总数,否则loss等于直接求和
  optional bool normalize = 2;
}

三、参考文献

<深度学习21天实战caffe>
Caffe源码解析(一)——caffe.proto