Net類是Caffe中的一個核心類, 他將各個Layer和Blob組織到一起, 完成前向和反響傳播計算

下面詳細解讀Net.cpp的程式碼:

----------------------------------------------------------------------------------------------------------------------------------

Net<Dtype>::Net 建構函式，在訓練一個網路時，涉及到多次迭代的設定和引數調優的策略，往往由更上一層的Solver來構造生成Net； 在用一個網路進行預測時，由於只需操作簡單的前向計算，不需要引數調優，往往不使用Solver，直接由自己寫的程式碼來構造Net 

//建構函式。NetParameter是在Solver構造Net時傳入的網路結構引數
//本建構函式一般在訓練階段由Solver呼叫
//root_net是根net, 如果本net非根net, 可以從root_net中複製layer
template <typename Dtype>
Net<Dtype>::Net(const NetParameter& param, const Net* root_net)
    : root_net_(root_net) {
  Init(param);
}

//建構函式。通過prototxt檔案來傳入網路結構引數
//本建構函式一般在測試階段由開發者自己寫的程式碼來呼叫
//phase: 當前net是進行TEST還是TRAIN
//stage: 指明當前net中哪些layer要包含在net中（通過在prototxt為各個layer的include/exclude中指明stage/not_stage,來確定某個layer是否要包含）
//level: 指明當前net中哪些layer要包含在net中（通過在prototxt為各個layer的include/exclude中指明level,來確定某個layer是否要包含）
template <typename Dtype>
Net<Dtype>::Net(const string& param_file, Phase phase,
    const int level, const vector<string>* stages,
    const Net* root_net)
    : root_net_(root_net) {
  NetParameter param;
  ReadNetParamsFromTextFileOrDie(param_file, ¶m);
  // Set phase, stages and level
  param.mutable_state()->set_phase(phase);
  if (stages != NULL) {
    for (int i = 0; i < stages->size(); i++) {
	  //在NetState中新增stages, NetState會與NetStateRules（來自於prototxt的include/exclude）比較來確定某個layer是否保留
      param.mutable_state()->add_stage((*stages)[i]);
    }
  }
  //在NetState中新增level, NetState會與NetStateRules（來自於prototxt的include/exclude）比較來確定某個layer是否保留
  param.mutable_state()->set_level(level);
  Init(param);
}
 

Net<Dtype>::Init 對net中的各個layer、每個layer的輸入輸出blob、layer中的引數blob進行初始化 

template <typename Dtype>
void Net<Dtype>::Init(const NetParameter& in_param) {
  //檢測是否有根net
  CHECK(Caffe::root_solver() || root_net_)
      << "root_net_ needs to be set for all non-root solvers";
  // Set phase from the state.
  phase_ = in_param.state().phase();
  // Filter layers based on their include/exclude rules and
  // the current NetState.
  NetParameter filtered_param;
  //使用stages/level規則對NetParameter進行過濾, 過濾後放入filtered_param
  FilterNet(in_param, &filtered_param); 
  LOG_IF(INFO, Caffe::root_solver())
      << "Initializing net from parameters: " << std::endl
      << filtered_param.DebugString();
  // Create a copy of filtered_param with splits added where necessary.
  NetParameter param;
  //對過濾後的filtered_param拷一個副本
  InsertSplits(filtered_param, ¶m);
  // Basically, build all the layers and set up their connections.
  name_ = param.name();
  map<string, int> blob_name_to_idx;
  set<string> available_blobs;
  memory_used_ = 0;
  // For each layer, set up its input and output
  //各個層輸入的blob。bottom_vecs_是二維vecter, 
  //第一維是各個層,第二維是某層中的各個bottom_blob
  bottom_vecs_.resize(param.layer_size());
  //各個層輸出blob
  top_vecs_.resize(param.layer_size());
  //各個層輸入blob的id
  bottom_id_vecs_.resize(param.layer_size());
  //各個層內參數blob的id
  param_id_vecs_.resize(param.layer_size());
  //各個層輸出blob的id
  top_id_vecs_.resize(param.layer_size());
  //各個層輸入blob是否要反向傳播計算
  bottom_need_backward_.resize(param.layer_size());
  for (int layer_id = 0; layer_id < param.layer_size(); ++layer_id) {
    // For non-root solvers, whether this layer is shared from root_net_.
    //當前層是否要從root_solver中共享得到（如果當前是根net，值為0）
    bool share_from_root = !Caffe::root_solver()
        && root_net_->layers_[layer_id]->ShareInParallel();
    // Inherit phase from net if unset.
    if (!param.layer(layer_id).has_phase()) {
      param.mutable_layer(layer_id)->set_phase(phase_);
    }
    // Setup layer.
    const LayerParameter& layer_param = param.layer(layer_id);
    if (layer_param.propagate_down_size() > 0) {
      CHECK_EQ(layer_param.propagate_down_size(),
          layer_param.bottom_size())
          << "propagate_down param must be specified "
          << "either 0 or bottom_size times ";
    }
	
	
    //////////////////////////////////////////////////////////////
    //建立層, 壓入layers_: 可以從root_solver共享獲得, 或者新建層
    //////////////////////////////////////////////////////////////
    if (share_from_root) {
      LOG(INFO) << "Sharing layer " << layer_param.name() << " from root net";
      //root_net_->layers_中包含了所有的layer,
      layers_.push_back(root_net_->layers_[layer_id]); 
      layers_[layer_id]->SetShared(true);
    } else { 
      //對於root_solver()中的net: 裡面所有layer都要新create出來
      //對於非root_solver()中的net, 只有非共享的layer才用新create
      layers_.push_back(LayerRegistry<Dtype>::CreateLayer(layer_param));
    }
    layer_names_.push_back(layer_param.name());
    LOG_IF(INFO, Caffe::root_solver())
        << "Creating Layer " << layer_param.name();
    bool need_backward = false;

	
	
    
    /////////////////////////////////////////////////////////////////////////
    //1.初始化bottom blob: 將bottom_vecs_的地址與blobs_[blob_id]地址關聯起來,
    //將bottom_id_vecs_與blob_id_關聯起來;
    //2.對於資料輸入層來說只有top,沒有bottom,所以會跳過下面的for迴圈
    ////////////////////////////////////////////////////////////////////////
    for (int bottom_id = 0; bottom_id < layer_param.bottom_size();
         ++bottom_id) {
      //1.net中bottom/top是交替初始化的,前一層的top是後一層的bottom,前一層top的
      //available_blobs/blob_name_to_idx引數就是後一層的bottom引數
      //2.AppendBottom將bottom_vecs_與blobs_[id]關聯起來, 將bottom_id_vecs_與
      //blob_id_關聯起來
      const int blob_id = AppendBottom(param, layer_id, bottom_id,
                                       &available_blobs, &blob_name_to_idx);
      // If a blob needs backward, this layer should provide it.
      //blob_need_backward_[blob_id]的值是由前一層top_blob傳遞過來的，同時與當
      //前層bottom_need_backward_[layer_id][bottom_id]或運算出來的結果;
      //need_backward是當前層是否要做反向傳播計算的最終判斷: need_backward由
      //所有blob_need_backward_和param_need_backward_組合得到
      need_backward |= blob_need_backward_[blob_id]; 
    }
    int num_top = layer_param.top_size();
	
	
	
	
    ////////////////////////////////////////////////////////////////////////
    //初始化top blob: 將top_vecs_的地址與blobs_[blob_id]地址關聯起來,
    //將top_id_vecs_與blob_id_關聯起來; AppendTop還建立了新blob
    ////////////////////////////////////////////////////////////////////////
    for (int top_id = 0; top_id < num_top; ++top_id) {
      //通過AppendTop和AppendBottom, bottom_vecs_和top_vecs_連線在了一起
      //在AppendTop中會往available_blobs新增某層的輸出blob,在AppendBottom中會
      //從available_blobs中刪除前一層的輸出blob，所有layers遍歷完後剩下的就
      //是整個net的輸出blob
      AppendTop(param, layer_id, top_id, &available_blobs, &blob_name_to_idx);
      // Collect Input layer tops as Net inputs.
      if (layer_param.type() == "Input") {
	//對於整個net的輸入層，每通過AppendTop新建一個top blob, blobs.size()
	//就增加1,blobs_size()是從0開始增加的，就能代表整個net輸入blob的id
        const int blob_id = blobs_.size() - 1;
        net_input_blob_indices_.push_back(blob_id);
        net_input_blobs_.push_back(blobs_[blob_id].get());
      }
    }
    // If the layer specifies that AutoTopBlobs() -> true and the LayerParameter
    // specified fewer than the required number (as specified by
    // ExactNumTopBlobs() or MinTopBlobs()), allocate them here.
    Layer<Dtype>* layer = layers_[layer_id].get();
	
	
	
	
    ////////////////////////////////////////////////////////////////////////
    //補上top blob, 使該層的top blob個數達到要求
    ////////////////////////////////////////////////////////////////////////
    if (layer->AutoTopBlobs()) {
      const int needed_num_top =
          std::max(layer->MinTopBlobs(), layer->ExactNumTopBlobs());
      //只有噹噹前層已有的top blob個數(num_top)小於引數中定義的個
      //數（needed_num_top）時 ,才需要自動生成blobs,補上缺口
      for (; num_top < needed_num_top; ++num_top) {
        // Add "anonymous" top blobs -- do not modify available_blobs or
        // blob_name_to_idx as we don't want these blobs to be usable as input
        // to other layers.
        AppendTop(param, layer_id, num_top, NULL, NULL);
      }
    }
	
	
	
    //////////////////////////////////////////////////////////////////////////
    // After this layer is connected, set it up.
    // 初始化每一個top Blob的shape(前面已經把bottom blob和top blob地址關聯起來了
    //, 所以不需要對bottom blob進行shape)
    //////////////////////////////////////////////////////////////////////////
    if (share_from_root) {
      // Set up size of top blobs using root_net_
      const vector<Blob<Dtype>*>& base_top = root_net_->top_vecs_[layer_id];
      const vector<Blob<Dtype>*>& this_top = this->top_vecs_[layer_id];
      for (int top_id = 0; top_id < base_top.size(); ++top_id) {
        this_top[top_id]->ReshapeLike(*base_top[top_id]);
        LOG(INFO) << "Created top blob " << top_id << " (shape: "
            << this_top[top_id]->shape_string() <<  ") for shared layer "
            << layer_param.name();
      }
    } else { 
      //如果是caffe::root_solver, 或非caffe::root_solver的非共享層的, 都會走下面分支
      layers_[layer_id]->SetUp(bottom_vecs_[layer_id], top_vecs_[layer_id]);
    }
    LOG_IF(INFO, Caffe::root_solver())
        << "Setting up " << layer_names_[layer_id];
	
	
	
	
    /////////////////////////////////////////////////////////////////////
    //初始化blob_loss_weights_: blob_loss_weights_用於存放loss；
    //blob_loss_weights_覆蓋了所有層的top blob, 但只有最後一層
    //Loss輸出層值才是非0
    //////////////////////////////////////////////////////////////////////
    for (int top_id = 0; top_id < top_vecs_[layer_id].size(); ++top_id) {
      if (blob_loss_weights_.size() <= top_id_vecs_[layer_id][top_id]) {

        //top_id_vecs_[layer_id][top_id]是整個net中順序排列的id號(不是本層
        //中從0開始的序列號 );每一個top blob都對應一個blob_loss_weights_[id]
	//的值， 用來存放loss,除了整個net的輸出blob外，值都是0
        blob_loss_weights_.resize(top_id_vecs_[layer_id][top_id] + 1, Dtype(0));
      }
      blob_loss_weights_[top_id_vecs_[layer_id][top_id]] = layer->loss(top_id);
      LOG_IF(INFO, Caffe::root_solver())
          << "Top shape: " << top_vecs_[layer_id][top_id]->shape_string();
      if (layer->loss(top_id)) {
        LOG_IF(INFO, Caffe::root_solver())
            << "    with loss weight " << layer->loss(top_id);
      }
      memory_used_ += top_vecs_[layer_id][top_id]->count();
    }
    LOG_IF(INFO, Caffe::root_solver())
        << "Memory required for data: " << memory_used_ * sizeof(Dtype);
	
	
	
    ///////////////////////////////////////////////////////////////////////
    //對引數進行初始化：一般權值weight存放在一個blob,偏執bias存放在另一個blob
    //本層的param_need_backward(具體值來自LayerParameter)和本層的
    //blob_need_backward_決定了本層的need_backward；本層的need_backward決
    //定了本層的layer_need_backward_
    ///////////////////////////////////////////////////////////////////////
    //LayerParameter中已經定義了的引數個數(可能小於實際的個數 )
    const int param_size = layer_param.param_size();
    //某層的實際引數個數
    const int num_param_blobs = layers_[layer_id]->blobs().size();
    CHECK_LE(param_size, num_param_blobs)
        << "Too many params specified for layer " << layer_param.name();
    ParamSpec default_param_spec;
    for (int param_id = 0; param_id < num_param_blobs; ++param_id) {
      const ParamSpec* param_spec = (param_id < param_size) ?
          &layer_param.param(param_id) : &default_param_spec;
      //lr_mult是收斂速率
      const bool param_need_backward = param_spec->lr_mult() != 0; 
      //need_backward是當前層是否要做反向傳播計算的最終判斷: need_backward由
      //所有blob_need_backward_和param_need_backward_組合得到
      need_backward |= param_need_backward;
      layers_[layer_id]->set_param_propagate_down(param_id,
                                                  param_need_backward);
    }
	
    //一個layer一般有兩個引數Blob, 第一個存weight, 第二個存bias
    for (int param_id = 0; param_id < num_param_blobs; ++param_id) {
      AppendParam(param, layer_id, param_id);
    }
    // Finally, set the backward flag
    // 只要本層中所有bottom blob和所有param blob中有一個支援backward, 
    //need_backward就為true
    layer_need_backward_.push_back(need_backward);
    if (need_backward) {
      //當本層要支援backward後， 本層所有blob都要支援backward
      for (int top_id = 0; top_id < top_id_vecs_[layer_id].size(); ++top_id) {
        //對top blob對應id的blob_need_backward_置true， 該結果會傳遞到後面一層
	//的bottom blob
        blob_need_backward_[top_id_vecs_[layer_id][top_id]] = true;
      }
    }
  }
  // Go through the net backwards to determine which blobs contribute to the
  // loss.  We can skip backward computation for blobs that don't contribute
  // to the loss.
  // Also checks if all bottom blobs don't need backward computation (possible
  // because the skip_propagate_down param) and so we can skip bacward
  // computation for the entire layer
  set<string> blobs_under_loss;
  set<string> blobs_skip_backp;
  
  //for迴圈遍歷每個layer, 將不需要backward計算的層和bottom_blob標記出來
  for (int layer_id = layers_.size() - 1; layer_id >= 0; --layer_id) {
    bool layer_contributes_loss = false;
    bool layer_skip_propagate_down = true;
	
    ///////////////////////////////////////////////////////////////////
    //遍歷該層每個top_blob,  確定該層是否輸出loss, 是否要backward計算
    ///////////////////////////////////////////////////////////////////
    for (int top_id = 0; top_id < top_vecs_[layer_id].size(); ++top_id) {
      const string& blob_name = blob_names_[top_id_vecs_[layer_id][top_id]];

      //如果當前層是最終輸出層，或當前top blob為最終loss做出貢獻了就把
      //layer_contributes_loss置true,layer_contributes_loss的true值最開
      //始的源頭是整個net的最終輸出層，之後每一層的layer_contributes_loss
      //通過判斷是否有top blob在blobs_under_loss中得到，blobs_under_loss
      //的值是由上一層bottom計算時插入的
      if (layers_[layer_id]->loss(top_id) ||
          (blobs_under_loss.find(blob_name) != blobs_under_loss.end())) {
        layer_contributes_loss = true;
      }
      if (blobs_skip_backp.find(blob_name) == blobs_skip_backp.end()) {
        layer_skip_propagate_down = false;
      }
	  
      //只要一層中有一個blob貢獻了loss,有一個blob要backwards, 就得到了該層這
      //兩個引數的最終結果,可以直接退出迴圈
      if (layer_contributes_loss && !layer_skip_propagate_down)
        break;
    }
    // If this layer can skip backward computation, also all his bottom blobs
    // don't need backpropagation	
    //該層如果同時滿足下面if中兩個條件, 就相互矛盾, 該層就不進行backward計算
    if (layer_need_backward_[layer_id] && layer_skip_propagate_down) {
      layer_need_backward_[layer_id] = false;
      for (int bottom_id = 0; bottom_id < bottom_vecs_[layer_id].size();
               ++bottom_id) {
        bottom_need_backward_[layer_id][bottom_id] = false;
      }
    }
    if (!layer_contributes_loss) { layer_need_backward_[layer_id] = false; }
    if (Caffe::root_solver()) {
      if (layer_need_backward_[layer_id]) {
        LOG(INFO) << layer_names_[layer_id] << " needs backward computation.";
      } else {
        LOG(INFO) << layer_names_[layer_id]
            << " does not need backward computation.";
      }
    }
    for (int bottom_id = 0; bottom_id < bottom_vecs_[layer_id].size();
         ++bottom_id) {
      if (layer_contributes_loss) {
        const string& blob_name =
            blob_names_[bottom_id_vecs_[layer_id][bottom_id]];
	//插入blobs_under_loss
        blobs_under_loss.insert(blob_name);
      } else {
	//如果該層沒有為loss做出貢獻, 該層就不需要backward計算
        bottom_need_backward_[layer_id][bottom_id] = false; 
      }
      if (!bottom_need_backward_[layer_id][bottom_id]) {
        const string& blob_name =
                   blob_names_[bottom_id_vecs_[layer_id][bottom_id]];
        blobs_skip_backp.insert(blob_name);
      }
    }
  }
  
  //////////////////////////////////////////////////////////////////////////////
  //如果當前net需要force backward, 將layer_need_backward設成true
  //blob_need_backward會由 layers_[layer_id]->AllowForceBackward(bottom_id)決定
  //////////////////////////////////////////////////////////////////////////////
  if (param.force_backward()) {
    for (int layer_id = 0; layer_id < layers_.size(); ++layer_id) {
      layer_need_backward_[layer_id] = true;
      for (int bottom_id = 0;
           bottom_id < bottom_need_backward_[layer_id].size(); ++bottom_id) {
        bottom_need_backward_[layer_id][bottom_id] =
            bottom_need_backward_[layer_id][bottom_id] ||
            layers_[layer_id]->AllowForceBackward(bottom_id);
        blob_need_backward_[bottom_id_vecs_[layer_id][bottom_id]] =
            blob_need_backward_[bottom_id_vecs_[layer_id][bottom_id]] ||
            bottom_need_backward_[layer_id][bottom_id];
      }
      for (int param_id = 0; param_id < layers_[layer_id]->blobs().size();
           ++param_id) {
        layers_[layer_id]->set_param_propagate_down(param_id, true);
      }
    }
  }
  // In the end, all remaining blobs are considered output blobs.
  //在AppendBottom中已經將bottom blob從available_blobs中刪掉,最終只剩下最頂
  //層的top blob,就是輸出blob
  for (set<string>::iterator it = available_blobs.begin(); 
      it != available_blobs.end(); ++it) {
    LOG_IF(INFO, Caffe::root_solver())
        << "This network produces output " << *it;
    net_output_blobs_.push_back(blobs_[blob_name_to_idx[*it]].get());
    net_output_blob_indices_.push_back(blob_name_to_idx[*it]);
  }
  for (size_t blob_id = 0; blob_id < blob_names_.size(); ++blob_id) {
    blob_names_index_[blob_names_[blob_id]] = blob_id;
  }
  for (size_t layer_id = 0; layer_id < layer_names_.size(); ++layer_id) {
    layer_names_index_[layer_names_[layer_id]] = layer_id;
  }
  ShareWeights();
  debug_info_ = param.debug_info();
  LOG_IF(INFO, Caffe::root_solver()) << "Network initialization done.";
}
 

Net::FilterNet 對NetParameter引數進行過濾, 僅保留符合規則的layers引數

Net::StateMeetsRule引數過濾的過程

在定義net結構的prototxt檔案中往往會定義某層的include/exclude引數, 比如下面的例子。include表示如果在構造net時如果滿足include的條件，本層就包含在net中；exclude表示在構造net時如果滿足exclude條件，本層就不會包含在net中。prototxt的這個include/exclude引數被讀取後就是caffe.proto中的NetStateRule類，類中有phase、min_level、max_level、stage、not_stage 5個引數，這些就是過濾得規則。那拿什麼來和這個規則進行比較呢？是用構造net時的輸入引數，我們往往用如下的方法來構造一個新net： Net<Dtype>::Net(const string& param_file, Phase phase,   const int level, const vector<string>* stages,   const Net* root_net)
對於包含include引數的層：如果滿足min_level<level<max_level 或 stages中任意一個元素能在NetStateRule::stage中找到, 該層就會被保留在net中 對於包含exclude引數的層：如果滿足min_level<level<max_level  或 stages中任意一個元素能在NetStateRule::stage中找到, 該層就會從net中剔除 當然如果是在NetStateRule::not_stage中找到， 結果正好相反

layer {
  name: "mnist"
  type: "Data"
  top: "data"
  top: "label"
  include {
    phase: TEST
    not_stage: "predict"    # 在 predict 時過濾掉這一層
  }
  transform_param {
    scale: 0.00390625
  }
  data_param {
    source: "examples/mnist/mnist_test_lmdb"
    batch_size: 100
    backend: LMDB
  }
}
# 增加 deploy 的輸入層
layer {
  name: "data"
  type: "Input"
  top: "data"
  input_param { shape: { dim: 1 dim: 1 dim: 28 dim: 28 } }
  exclude {
    phase: TEST
    stage: "predict"    # 在 predict 時不加上這一層
  }
}

對引數進行過濾有什麼實際用處呢？ 可以參考這個文章：https://yangwenbo.com/articles/caffe-net-config-all-in-one.html?utm_source=tuicool&utm_medium=referral

template <typename Dtype>
void Net<Dtype>::FilterNet(const NetParameter& param,
    NetParameter* param_filtered) {
  
  NetState net_state(param.state());
  param_filtered->CopyFrom(param);
  //先清除layers,然後根據規則重新新增layers
  param_filtered->clear_layer();
  for (int i = 0; i < param.layer_size(); ++i) {
    const LayerParameter& layer_param = param.layer(i);
    const string& layer_name = layer_param.name();
	//include和exclude不能同時存在
    CHECK(layer_param.include_size() == 0 || layer_param.exclude_size() == 0)
          << "Specify either include rules or exclude rules; not both.";
    // If no include rules are specified, the layer is included by default and
    // only excluded if it meets one of the exclude rules.
    bool layer_included = (layer_param.include_size() == 0);
    for (int j = 0; layer_included && j < layer_param.exclude_size(); ++j) {
	  //net_state是由構造net時的輸入引數組成(phase/stage/level);
      //layer_param.exclude是在prototxt中某層的exclude中的參
	  //數(max_level/min_level/stage/not_stage/phase);
	  //滿足if條件就說明,本層要被exclude;
      if (StateMeetsRule(net_state, layer_param.exclude(j), layer_name)) {
        layer_included = false;
      }
    }
    for (int j = 0; !layer_included && j < layer_param.include_size(); ++j) {
	  //滿足條件就說明,本層要被include
      if (StateMeetsRule(net_state, layer_param.include(j), layer_name)) {
        layer_included = true;
      }
    }
    if (layer_included) {
      param_filtered->add_layer()->CopyFrom(layer_param);
    }
  }
}


//用構造net時的輸入phase/level/stage與prototxt中各層的規則(include/exclude)
//比較,決定本層是否要包含在net中
template <typename Dtype>
bool Net<Dtype>::StateMeetsRule(const NetState& state,
    const NetStateRule& rule, const string& layer_name) {
  // Check whether the rule is broken due to phase.
  if (rule.has_phase()) {
      if (rule.phase() != state.phase()) {
        LOG_IF(INFO, Caffe::root_solver())
            << "The NetState phase (" << state.phase()
            << ") differed from the phase (" << rule.phase()
            << ") specified by a rule in layer " << layer_name;
        return false;
      }
  }
  // Check whether the rule is broken due to min level.
  if (rule.has_min_level()) {
    if (state.level() < rule.min_level()) {
      LOG_IF(INFO, Caffe::root_solver())
          << "The NetState level (" << state.level()
          << ") is above the min_level (" << rule.min_level()
          << ") specified by a rule in layer " << layer_name;
      return false;
    }
  }
  // Check whether the rule is broken due to max level.
  if (rule.has_max_level()) {
    if (state.level() > rule.max_level()) {
      LOG_IF(INFO, Caffe::root_solver())
          << "The NetState level (" << state.level()
          << ") is above the max_level (" << rule.max_level()
          << ") specified by a rule in layer " << layer_name;
      return false;
    }
  }
  // Check whether the rule is broken due to stage. The NetState must
  // contain ALL of the rule's stages to meet it.
  for (int i = 0; i < rule.stage_size(); ++i) {
    // Check that the NetState contains the rule's ith stage.
    bool has_stage = false;
	//net構造時輸入的stage中只要有一個符合stage規則,就表明本層被include
    for (int j = 0; !has_stage && j < state.stage_size(); ++j) {
      if (rule.stage(i) == state.stage(j)) { has_stage = true; }
    }
    if (!has_stage) {
      LOG_IF(INFO, Caffe::root_solver())
          << "The NetState did not contain stage '" << rule.stage(i)
          << "' specified by a rule in layer " << layer_name;
      return false;
    }
  }
  // Check whether the rule is broken due to not_stage. The NetState must
  // contain NONE of the rule's not_stages to meet it.
  for (int i = 0; i < rule.not_stage_size(); ++i) {
    // Check that the NetState contains the rule's ith not_stage.
    bool has_stage = false;
	//net構造時輸入的stage中只要有一個符合not_stage規則,就表明本層被exclude
    for (int j = 0; !has_stage && j < state.stage_size(); ++j) {
      if (rule.not_stage(i) == state.stage(j)) { has_stage = true; }
    }
    if (has_stage) {
      LOG_IF(INFO, Caffe::root_solver())
          << "The NetState contained a not_stage '" << rule.not_stage(i)
          << "' specified by a rule in layer " << layer_name;
      return false;
    }
  }
  return true;
}

Net::AppendTop 給某層增加一個top blob

Net::AppendBottom給某層增加一個bottom blob

AppendTop函式會向整個net的blob列表（blobs_）中新增一個新blob，同時將本層新建的top blob指向該新增blob， 這樣就把層的輸出blob和blob列表(blobs_)關聯起來了。AppendTop函式在新建blob時可能會採用同址計算（in-place computer），所謂同址計算就是同一層的top blob和bottom blob複用。 AppendBottom函式不會向blobs_新增blob了，只是簡單的把新增的bottom blob和在AppendTop中已經增加的blobs_關聯起來。 經過上述兩個函式的處理，前一層的top blob、當前層的bottom blob就通過blobs_關聯起來了，整個net中所有的層級就連結到一起。

// Helper for Net::Init: add a new top blob to the net.
// 給某層增加一個top blob
template <typename Dtype>
void Net<Dtype>::AppendTop(const NetParameter& param, const int layer_id,
                           const int top_id, set<string>* available_blobs,
                           map<string, int>* blob_name_to_idx) {
  shared_ptr<LayerParameter> layer_param(
      new LayerParameter(param.layer(layer_id)));
  const string& blob_name = (layer_param->top_size() > top_id) ?
      layer_param->top(top_id) : "(automatic)";
  // Check if we are doing in-place computation
  
  
  //同址計算:top blob使用和bottom blob相同的地址和id
  //是否使用同址計算由prototxt中對top/bottom blob名字的定義決定
  if (blob_name_to_idx && layer_param->bottom_size() > top_id &&
      blob_name == layer_param->bottom(top_id)) {
    // In-place computation	
    LOG_IF(INFO, Caffe::root_solver())
        << layer_param->name() << " -> " << blob_name << " (in-place)";
    top_vecs_[layer_id].push_back(blobs_[(*blob_name_to_idx)[blob_name]].get());
    top_id_vecs_[layer_id].push_back((*blob_name_to_idx)[blob_name]);
  } else if (blob_name_to_idx &&
             blob_name_to_idx->find(blob_name) != blob_name_to_idx->end()) {
	//blob_name_to_idx中的元素始終是在AppendTop中新增的,所以如果有重複名字,
	//就意味之前有其他top blob同名
    // If we are not doing in-place computation but have duplicated blobs,
    // raise an error.
    LOG(FATAL) << "Top blob '" << blob_name
               << "' produced by multiple sources.";
  } else {
    // Normal output.
	// 不進行同址計算, top使用和bottom獨立的Blob
    if (Caffe::root_solver()) {
      LOG(INFO) << layer_param->name() << " -> " << blob_name;
    }
    shared_ptr<Blob<Dtype> > blob_pointer(new Blob<Dtype>());
	//當前blob的個數, 就是要新增的blob的id(在 blobs_尾部新增一個blob)
    const int blob_id = blobs_.size();
    //新增一個blob的動作是在AppendTop中完成的, AppendBottom中只是把當前層bottom和
	//前一層top的地址關聯起來(通過bottom/top指向相同的blobs_[id]/blob_id來連線)
	blobs_.push_back(blob_pointer);
    blob_names_.push_back(blob_name);
    blob_need_backward_.push_back(false);
	//新增blob_name_to_idx的鍵值, 對於資料輸入層(只有top,沒有bottom, 是第一層),
	//blob_name_to_idx也不是null,所以也會進入下面分支
    if (blob_name_to_idx) { (*blob_name_to_idx)[blob_name] = blob_id; }
    top_id_vecs_[layer_id].push_back(blob_id);
    top_vecs_[layer_id].push_back(blob_pointer.get());
  }
  //在AppendTop中增加blob,在AppendBottom中剔除blob,遍歷所有層後剩下的就是net的輸出blob
  if (available_blobs) { available_blobs->insert(blob_name); }
}

// Helper for Net::Init: add a new bottom blob to the net.
// 給某層增加一個bottom blob
// 將bottom_vecs_與blobs_[id]關聯起來, 將bottom_id_vecs_與blob_id_關聯起來
template <typename Dtype>
int Net<Dtype>::AppendBottom(const NetParameter& param, const int layer_id,
    const int bottom_id, set<string>* available_blobs,
    map<string, int>* blob_name_to_idx) {
  const LayerParameter& layer_param = param.layer(layer_id);
  const string& blob_name = layer_param.bottom(bottom_id);
  if (available_blobs->find(blob_name) == available_blobs->end()) {
    LOG(FATAL) << "Unknown bottom blob '" << blob_name << "' (layer '"
               << layer_param.name() << "', bottom index " << bottom_id << ")";
  }
  const int blob_id = (*blob_name_to_idx)[blob_name];
  LOG_IF(INFO, Caffe::root_solver())
      << layer_names_[layer_id] << " <- " << blob_name;
  //新增一個blob的動作是在top中完成的, bottom中只是把當前層bottom和前一層
  //top的地址連線起來(通過bottom/top指向相同的blobs_[id]/blob_id來連線)
  bottom_vecs_[layer_id].push_back(blobs_[blob_id].get());
  bottom_id_vecs_[layer_id].push_back(blob_id);
  //上一層的AppendTop時insert入available_blob, 本層的AppendBottom時erase
  available_blobs->erase(blob_name); 
  bool need_backward = blob_need_backward_[blob_id];
  // Check if the backpropagation on bottom_id should be skipped
  if (layer_param.propagate_down_size() > 0) {
    need_backward = layer_param.propagate_down(bottom_id);
  }
  bottom_need_backward_[layer_id].push_back(need_backward);
  return blob_id;
}

Net::AppendParam 給某層增加可學習引數blob和超訓練引數

可學習引數: 存放入learnable_params_中, 用於存放權重weight和偏置bias, 對於需要反響傳播計算的層, 至少有兩個引數blob(一個放權重, 一個放偏置)

超訓練引數: 存放入params_lr和params_weight_decay中, 用於存放收斂速率和權重衰減

每一個可學習引數的blob(不管是存放權重還是偏置)都對應有超訓練引數params_lr和params_weight_decay, 這三者都通過param_names_index來統一索引

/////////////////////////////////////////////////////////////////////////////////////
//1.給某層增加一個可學習引數blob(存放權重/偏置),放入params_, 同時放入learnable_params_;
//2.給某層增加一個params_lr_和params_weight_decay_,用來存放超訓練引數
//3.一層中每個可學習引數blob(權重/偏置, learnable_params_)都對應有一個params_lr_和一個
//params_weight_decay_, 超訓練引數和可學習引數都是從LayerParameter中獲取到.
//4.param_names_index_是對一層中可學習引數/超訓練引數的總索引
////////////////////////////////////////////////////////////////////////////////////
template <typename Dtype>
void Net<Dtype>::AppendParam(const NetParameter& param, const int layer_id,
                             const int param_id) {
  //layers_[layer_id]->layer_param()中存放的是訓練超引數
  const LayerParameter& layer_param = layers_[layer_id]->layer_param();
  const int param_size = layer_param.param_size();
  string param_name =
      (param_size > param_id) ? layer_param.param(param_id).name() : "";
  if (param_name.size()) {
    param_display_names_.push_back(param_name);
  } else {
    ostringstream param_display_name;
    param_display_name << param_id;
    param_display_names_.push_back(param_display_name.str());
  }
  const int net_param_id = params_.size();
  //layers_[layer_id]->blobs()中存放的是可學習引數(權重/偏置);
  //一個層一般有兩個blob,第一個存weight,第二個存bias
  params_.push_back(layers_[layer_id]->blobs()[param_id]);
  param_id_vecs_[layer_id].push_back(net_param_id);
  param_layer_indices_.push_back(make_pair(layer_id, param_id));
  //param_spec用來存放某層的訓練超引數
  ParamSpec default_param_spec;
  const ParamSpec* param_spec = (layer_param.param_size() > param_id) ?
      &layer_param.param(param_id) : &default_param_spec;
  if (!param_size || !param_name.size() || (param_name.size() &&
      param_names_index_.find(param_name) == param_names_index_.end())) {
    // This layer "owns" this parameter blob -- it is either anonymous
    // (i.e., not given a param_name) or explicitly given a name that we
    // haven't already seen.
    //param_owners_用於存放param_names_index_中的id值, 如果是新建的
    //param_names_index_引數,id為-1
    param_owners_.push_back(-1);
    if (param_name.size()) {
      param_names_index_[param_name] = net_param_id;
    }
    const int learnable_param_id = learnable_params_.size();
    //learnable_params_存放權重/偏置blob
    learnable_params_.push_back(params_[net_param_id].get());
    //learnable_param_ids_存放權重/偏置blob的id
    learnable_param_ids_.push_back(learnable_param_id);
    //has_params_lr_存放收斂速率的開關
    has_params_lr_.push_back(param_spec->has_lr_mult());
    //has_params_decay_存放權重衰減的開關
    has_params_decay_.push_back(param_spec->has_decay_mult());
    //params_lr_存放收斂速率值
    params_lr_.push_back(param_spec->lr_mult());
    //params_weight_decay_存放權重衰減值
    params_weight_decay_.push_back(param_spec->decay_mult());
  } else {
    // Named param blob with name we've seen before: share params
	// 複用之前已存在param_names_index_表中的引數,
	// param_names_index_是對一層中可學習引數/超訓練引數的總索引
    const int owner_net_param_id = param_names_index_[param_name];
    param_owners_.push_back(owner_net_param_id);//param_owners_用於存放param_names_index_中的id值, 如果是新建的param_names_index_引數,id為-1
    const pair<int, int>& owner_index =
        param_layer_indices_[owner_net_param_id];
    const int owner_layer_id = owner_index.first;
    const int owner_param_id = owner_index.second;
    LOG_IF(INFO, Caffe::root_solver()) << "Sharing parameters '" << param_name
        << "' owned by "
        << "layer '" << layer_names_[owner_layer_id] << "', param "
        << "index " << owner_param_id;
    Blob<Dtype>* this_blob = layers_[layer_id]->blobs()[param_id].get();
    Blob<Dtype>* owner_blob =
        layers_[owner_layer_id]->blobs()[owner_param_id].get();
    const int param_size = layer_param.param_size();
    if (param_size > param_id && (layer_param.param(param_id).share_mode() ==
                                  ParamSpec_DimCheckMode_PERMISSIVE)) {
      // Permissive dimension checking -- only check counts are the same.
      CHECK_EQ(this_blob->count(), owner_blob->count())
          << "Cannot share param '" << param_name << "' owned by layer '"
          << layer_names_[owner_layer_id] << "' with layer '"
          << layer_names_[layer_id] << "'; count mismatch.  Owner layer param "
          << "shape is " << owner_blob->shape_string() << "; sharing layer "
          << "shape is " << this_blob->shape_string();
    } else {
      // Strict dimension checking -- all dims must be the same.
      CHECK(this_blob->shape() == owner_blob->shape())
          << "Cannot share param '" << param_name << "' owned by layer '"
          << layer_names_[owner_layer_id] << "' with layer '"
          << layer_names_[layer_id] << "'; shape mismatch.  Owner layer param "
          << "shape is " << owner_blob->shape_string() << "; sharing layer "
          << "expects shape " << this_blob->shape_string();
    }
    //因為複用, 需要把learnable_param_id再重複放入learnable_param_ids_中
    const int learnable_param_id = learnable_param_ids_[owner_net_param_id];
    learnable_param_ids_.push_back(learnable_param_id);
    if (param_spec->has_lr_mult()) {
      if (has_params_lr_[learnable_param_id]) {
        CHECK_EQ(param_spec->lr_mult(), params_lr_[learnable_param_id])
            << "Shared param '" << param_name << "' has mismatched lr_mult.";
      } else {
        has_params_lr_[learnable_param_id] = true;
        params_lr_[learnable_param_id] = param_spec->lr_mult();
      }
    }
    if (param_spec->has_decay_mult()) {
      if (has_params_decay_[learnable_param_id]) {
        CHECK_EQ(param_spec->decay_mult(),
                 params_weight_decay_[learnable_param_id])
            << "Shared param '" << param_name << "' has mismatched decay_mult.";
      } else {
        has_params_decay_[learnable_param_id] = true;
        params_weight_decay_[learnable_param_id] = param_spec->decay_mult();
      }
    }
  }
}