net_->ForwardBackward()的大致梳理

net_->ForwardBackward()方法在net.hpp文件中

Dtype ForwardBackward() {
    Dtype loss;
    Forward(&loss);
    Backward();
    return loss;
  }

首先进入Forward(&loss)

net.cpp

template <typename Dtype>
const vector<Blob<Dtype>*>& Net<Dtype>::Forward(Dtype* loss) {
  if (loss != NULL) {
    *loss = ForwardFromTo(0, layers_.size() - 1);
  } else {
    ForwardFromTo(0, layers_.size() - 1);
  }
  return net_output_blobs_;
}

接着进入*loss = ForwardFromTo(0, layers_.size() - 1)这句话

net.cpp

template <typename Dtype>
Dtype Net<Dtype>::ForwardFromTo(int start, int end) {
  CHECK_GE(start, 0);
  CHECK_LT(end, layers_.size());
  Dtype loss = 0;
  for (int i = start; i <= end; ++i) {
    for (int c = 0; c < before_forward_.size(); ++c) {
      before_forward_[c]->run(i);
    }
    // 一层一层地前向传播，bottom_vecs_[i]是各层的输入输入数据指针，top_vecs_[i]是各层的输出数据指针
    Dtype layer_loss = layers_[i]->Forward(bottom_vecs_[i], top_vecs_[i]);
    loss += layer_loss;
    if (debug_info_) { ForwardDebugInfo(i); }
    for (int c = 0; c < after_forward_.size(); ++c) {
      after_forward_[c]->run(i);
    }
  }
  return loss;
}

再进入Dtype layer_loss = layers_[i]->Forward(bottom_vecs_[i], top_vecs_[i])。首先会进入Layer类的Forward函数

layer.hpp

// Forward and backward wrappers. You should implement the cpu and
// gpu specific implementations instead, and should not change these
// functions.
template <typename Dtype>
inline Dtype Layer<Dtype>::Forward(const vector<Blob<Dtype>*>& bottom,
    const vector<Blob<Dtype>*>& top) {
  Dtype loss = 0;
  Reshape(bottom, top);
  switch (Caffe::mode()) {
  case Caffe::CPU:
    // Layer类的虚函数，具体由其不同的派生类作不同的实现，也就是此句将会调用不同网络层的Forward_cpu函数，下面的Forward_gpu同理。
    Forward_cpu(bottom, top);
    for (int top_id = 0; top_id < top.size(); ++top_id) {
      if (!this->loss(top_id)) { continue; }
      const int count = top[top_id]->count();
      const Dtype* data = top[top_id]->cpu_data();
      const Dtype* loss_weights = top[top_id]->cpu_diff();
      loss += caffe_cpu_dot(count, data, loss_weights);
    }
    break;
  case Caffe::GPU:
    Forward_gpu(bottom, top);
#ifndef CPU_ONLY
    for (int top_id = 0; top_id < top.size(); ++top_id) {
      if (!this->loss(top_id)) { continue; }
      const int count = top[top_id]->count();
      const Dtype* data = top[top_id]->gpu_data();
      const Dtype* loss_weights = top[top_id]->gpu_diff();
      Dtype blob_loss = 0;
      caffe_gpu_dot(count, data, loss_weights, &blob_loss);
      loss += blob_loss;
    }
#endif
    break;
  default:
    LOG(FATAL) << "Unknown caffe mode.";
  }
  return loss;
}

template <typename Dtype>
inline void Layer<Dtype>::Backward(const vector<Blob<Dtype>*>& top,
    const vector<bool>& propagate_down,
    const vector<Blob<Dtype>*>& bottom) {
  switch (Caffe::mode()) {
  case Caffe::CPU:
    Backward_cpu(top, propagate_down, bottom);
    break;
  case Caffe::GPU:
    Backward_gpu(top, propagate_down, bottom);
    break;
  default:
    LOG(FATAL) << "Unknown caffe mode.";
  }
}

接下来再看ForwardBackward()中的Backward()

net.cpp

template <typename Dtype>
void Net<Dtype>::Backward() {
  // 从最后一层开始反向传播
  BackwardFromTo(layers_.size() - 1, 0);
  if (debug_info_) {
    Dtype asum_data = 0, asum_diff = 0, sumsq_data = 0, sumsq_diff = 0;
    for (int i = 0; i < learnable_params_.size(); ++i) {
      asum_data += learnable_params_[i]->asum_data();
      asum_diff += learnable_params_[i]->asum_diff();
      sumsq_data += learnable_params_[i]->sumsq_data();
      sumsq_diff += learnable_params_[i]->sumsq_diff();
    }
    const Dtype l2norm_data = std::sqrt(sumsq_data);
    const Dtype l2norm_diff = std::sqrt(sumsq_diff);
    LOG(ERROR) << "    [Backward] All net params (data, diff): "
               << "L1 norm = (" << asum_data << ", " << asum_diff << "); "
               << "L2 norm = (" << l2norm_data << ", " << l2norm_diff << ")";
  }
}

进入BackwardFromTo(layers_.size() - 1, 0)

net.cpp

template <typename Dtype>
void Net<Dtype>::BackwardFromTo(int start, int end) {
  CHECK_GE(end, 0);
  CHECK_LT(start, layers_.size());
  for (int i = start; i >= end; --i) {
    for (int c = 0; c < before_backward_.size(); ++c) {
      before_backward_[c]->run(i);
    }
    if (layer_need_backward_[i]) {
      // 反向传播过程中，top_vecs_[i]是各层的输入数据指针，bottom_vecs[i]是各层的输出数据指针，与前向传播正好相反
      layers_[i]->Backward(
          top_vecs_[i], bottom_need_backward_[i], bottom_vecs_[i]);
      if (debug_info_) { BackwardDebugInfo(i); }
    }
    for (int c = 0; c < after_backward_.size(); ++c) {
      after_backward_[c]->run(i);
    }
  }
}

进入layers_[i]->Backward(top_vecs_[i], bottom_need_backward_[i], bottom_vecs_[i])

layer.hpp

template <typename Dtype>
inline void Layer<Dtype>::Backward(const vector<Blob<Dtype>*>& top,
    const vector<bool>& propagate_down,
    const vector<Blob<Dtype>*>& bottom) {
  switch (Caffe::mode()) {
  case Caffe::CPU:
    // 与前向传播类似，利用不同派生类的同名函数作出不同层的反向传播的具体实现
    Backward_cpu(top, propagate_down, bottom);
    break;
  case Caffe::GPU:
    Backward_gpu(top, propagate_down, bottom);
    break;
  default:
    LOG(FATAL) << "Unknown caffe mode.";
  }
}

不同层的前向、反向传播的具体实现见下一章节。