caffe源码阅读
结构
主要两个目录
src: 包含源码实现
include: 头文件
src目录的架构,主要代码在caffe目录中,包含net.cpp, solver.cpp, blob.cpp, layer.cpp, blob.cpp, common.cpp, layers目录主要包含一些层,是caffe核心。proto中只有一个caffe.proto文件,里面使用protobuf语言描述了各种对象的成员变量, solvers主要提供不同的优化器,sgd, adam, rmsprop, adagrad,test目录包含一些单元测试用例, util常用工具函数:
├── caffe
│ ├── layers
│ ├── proto
│ ├── solvers
│ ├── test
│ │ └── test_data
│ └── util
└── gtest
首先来看caffe目录下的几个cpp:
blob.cpp
common.cpp
data_transformer.cpp
internal_thread.cpp
layer.cpp
layer_factory.cpp
net.cpp
parallel.cpp
solver.cpp
syncedmem.cpp
blob.cpp是caffe中主要的数据传输类型。
common.cpp
从tools出发
在根目录下有一个tools目录,主要用来编译一个caffe的可执行档,里面提供了caffe的一些可执行参数,通过配置参数来达到使用caffe的目的。
caffe.cpp
compute_image_mean.cpp
convert_imageset.cpp
device_query.cpp
extract_features.cpp
finetune_net.cpp
net_speed_benchmark.cpp
test_net.cpp
train_net.cpp
upgrade_net_proto_binary.cpp
upgrade_net_proto_text.cpp
upgrade_solver_proto_text.cpp
main.cpp
中分别注册了几个函数到g_brew_map中,分别是train, test, time, device_query。
首先来看train函数,使用一个solver_param对象来解析solver参数,
caffe::SolverParameter solver_param;
caffe::ReadSolverParamsFromTextFileOrDie(FLAGS_solver, &solver_param);
通过SolverRegistery::CreateSolver
创建一个solver对象, solver对象有一个 shared_ptr<Net<Dtype> > net_
成员变量:
shared_ptr<caffe::Solver<float> >
solver(caffe::SolverRegistry<float>::CreateSolver(solver_param));
Net
对象是整个网络的主体,那么一个Net
究竟包含什么呢?最主要的是三个变量, layers_
, params_
, blobs_
,如下:
template <typename Dtype>
class Net {
private:
vector<shared_ptr<Layer<Dtype> > > layers_;
vector<shared_ptr<Blob<Dtype> > > params_;
vector<shared_ptr<Blob<Dtype> > > blobs_;
};
layers_
是构成网络的基本组件; params_
是每层的滤波器参数,这个变量和每层layer
的blobs_
变量是共享数据的,即这边的params_
存储的是layer
的blobs_
的指针; blobs_
是各层的中间数据。
Net
构造函数接收一个NetParameter
参数,只是调用了一下Init
函数:
template <typename Dtype>
Net<Dtype>::Net(const NetParameter& param) {
Init(param);
}
NetParameter
在caffe.proto的定义如下:
message NetParameter {
optional string name = 1;
repeated string input = 3;
repeated BlobShape input_shape = 8;
repeated int32 input_dim = 4;
optional bool force_backward = 5 [default = false];
optional NetState state = 6;
repeated LayerParameter layer = 100; // ID 100 so layers are printed last.
}
message LayerParameter {
optional string name = 1; // the layer name
optional string type = 2; // the layer type
repeated string bottom = 3; // the name of each bottom blob
repeated string top = 4; // the name of each top blob
// The blobs containing the numeric parameters of the layer.
repeated BlobProto blobs = 7;
optional TransformationParameter transform_param = 100;
}
NetParameter
的核心是LayerParameter
,
LayerParamter
(定义进行了简化)的核心是bottom
名, top
名, 以及参数blobs
。
这个NetParamter
利用protobuf
从train.prototxt
, vgg.caffemodel
进行读取初始化,然后去构造Net
对象,有了Net
整个网络也就搭建起来了。
之后可以调用solver->Solve();
函数来开始整个网络的训练,而在Solve()
函数中,则调用Step()
函数,Step()
函数主要用来进行每次的迭代,里面有个循环,每个循环是一次iter,每个iter进行iter_size
次前向反向传播(FowardBackward()
),并对这个batch的loss取平均更新优化器。
这里的iter_size
参数是为了防止由于GPU内存不足导致无法使用较大的batch size
带来的问题,因为它实际更新loss的迭代次数是iter_size * batch_size
,这样就可以与使用较大的batch size
是相同的结果。例如网络在batch_size = 128
时取得较好的结果,但由于GPU内存不够,只够32张图片,那么可以将batch_size
设为32,将iter_size
设为4,取得的效果与batch_size = 128
一样。
while (iter_ < stop_iter) {
// ...
Dtype loss = 0;
for (int i = 0; i < param_.iter_size(); ++i) {
loss += net_->ForwardBackward();
}
loss /= param_.iter_size();
// average the loss across iterations for smoothed reporting
UpdateSmoothedLoss(loss, start_iter, average_loss);
// ...
ApplyUpdate();
// ...
}
查看FowardBackward()
实现如下,分别进行了Forward, Backward,并在前向传播时记录了loss:
Dtype ForwardBackward() {
Dtype loss;
Forward(&loss);
Backward();
return loss;
}
再看Foward(&loss)
实现,调用了FowardFromTo(0, layers_.size() - 1)
函数:
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_;
}
FowardFromTo(0, layers_.szie()-1)
遍历了每个层,使每个层分别调用Forward()
函数,bottom_vecs_
,top_vecs_
的类型是vector<vector<Blob<Dtype>*> >
,传入每层的类型是vector<Blob<Dtype>*>
,这个vector
表示层可能有多个输入或输出:
template <typename Dtype>
Dtype Net<Dtype>::ForwardFromTo(int start, int end) {
Dtype loss = 0;
for (int i = start; i <= end; ++i) {
Dtype layer_loss = layers_[i]->Forward(bottom_vecs_[i], top_vecs_[i]);
loss += layer_loss;
}
return loss;
}
所以,以上的solver, net都是为了layer服务,核心的功能实现还是在layer当中,我们先来看卷积层(conv_layer.cpp)的Forward实现。
LayerFactory: 工厂模式
为了对layer有足够的理解,我们先来阅读与layer相关的对象。所有layer的基类是Layer
,由于实现的类都是使用模板编程,如果没有静态地调用相关模板类,编译器是不会进行特化的。而我们的调用过程都是通过配置文件train.prototxt进行动态初始化相关的类,这样就会发现找不到这个类。为了避免这个问题,在类定义后面都进行一下声明,这样确保在使用的时候可以找到这个类,使用的是一个宏:
INSTANTIATE_CLASS(ConvolutionLayer);
宏的定义如下:
#define INSTANTIATE_CLASS(classname)
char gInstantiationGuard##classname;
template class classname<float>;
template class classname<double>
实际上就是声明了一下ConvolutionLayer<float>
, ConvolutionLayer<double>
:
char gInstantiationGuardConvolutionLayer;
template class ConvolutionLayer<float>;
template class ConvolutionLayer<double>;
除此之外,有那么多的Layer,caffe实现了一个工厂模型(layer_factory.cpp),将layer进行统一管理,也就是需要将所有Layer都注册到一个map,里面的key对应Layer名,value是生成相应的Layer函数,这样在使用的时候就可以根据类型实例化相应的Layer对象了。提供了两个宏定义:
#define REGISTER_LAYER_CREATOR(type, creator)
static LayerRegisterer<float> g_creator_f_##type(#type, creator<float>);
static LayerRegisterer<double> g_creator_d_##type(#type, creator<double>)
#define REGISTER_LAYER_CLASS(type)
template <typename Dtype>
shared_ptr<Layer<Dtype> > Creator_##type##Layer(const LayerParameter& param)
{
return shared_ptr<Layer<Dtype> >(new type##Layer<Dtype>(param));
}
REGISTER_LAYER_CREATOR(type, Creator_##type##Layer)
先看第一个宏,传入两个参数,一个是类型(Convolution),第二个是创建函数,如在layer_factory.cpp
中有如下代码(进行了简化):
template <typename Dtype>
shared_ptr<Layer<Dtype> > GetConvolutionLayer(const LayerParameter& param) {
// 简化...
return shared_ptr<Layer<Dtype> >(new ConvolutionLayer<Dtype>(param));
// 简化...
}
REGISTER_LAYER_CREATOR(Convolution, GetConvolutionLayer);
那么宏翻译过来就是如下:
static LayerRegisterer<float> g_creator_f_Convolution("Convolution", creator<float>);
static LayerRegisterer<double> g_creator_d_Convolution("Convolution", creator<double>) ;
所以我们再来看看LayerRegisterer
这个类干了什么:
LayerRegistry<Dtype>::AddCreator(type, creator);
调用了静态函数LayerRegistry<Dtype>::AddCreator
,继续看:
class LayerRegistry {
public:
static CreatorRegistry& Registry() {
static CreatorRegistry* g_registry_ = new map<string, Creator>();
return *g_registry_;
}
static void AddCreator(const string& type, Creator creator) {
CreatorRegistry& registry = Registry();
CHECK_EQ(registry.count(type), 0) << "Layer type " << type << " already registered.";
registry[type] = creator;
}
}
可以看到维护了一个单例map类型对象g_registry_
,这个对象存储了类型与对应的创建函数。
第二个宏,假如是这样调用REGISTER_LAYER_CLASS(Convolution),则可以翻译成下面的样子:
template <typename Dtype>
shared_ptr<Layer<Dtype> > Creator_ConvolutionLayer(const LayerParameter& param)
{
return shared_ptr<Layer<Dtype> >(new ConvolutionLayer<Dtype>(param));
}
REGISTER_LAYER_CREATOR(type, Creator_ConvolutionLayer)
就是这个类不需要特殊创建,直接使用这个默认创建方法(Creator_ConvolutionLayer
)就可以。而一些特殊的例子比如Convolution要进行其它的处理,所以要特殊写创建函数(GetConvolutionLayer
),当然大多数层都可以直接调用这个默认的函数进行创建。
数据Blob
caffe中的数据的基本存储、操作对象就是Blob,还提供了CPU、GPU数据同步功能。
Blob的数据基本存储就是数组,是按照行存储的。
Blob主要存储了两个数据,data_, diff_,分别是数据与梯度。
blob是一个四维的数组。维度从高到低分别是:(num_,channels_,height_,width_)
对于图像数据来说就是:图片个数,彩色通道个数,宽,高,比如说有10张图片,分别是512*256
大小,彩色三通道,则为:(10,3,256,512)
:
template <typename Blob>
class Blob {
public:
inline int num() const { return LegacyShape(0); }
inline int channels() const { return LegacyShape(1); }
inline int height() const { return LegacyShape(2); }
inline int width() const { return LegacyShape(3); }
inline const shared_ptr<SyncedMemory>& data() const {
return data_;
}
inline const shared_ptr<SyncedMemory>& diff() const {
return diff_;
}
void Update() {
caffe_axpy<Dtype>(count_, Dtype(-1), static_cast<const Dtype*>(diff_->cpu_data()), static_cast<Dtype*>(data_->mutable_cpu_data()));
}; // 数据更新,即减去当前计算出来的梯度
void FromProto(const BlobProto& proto, bool reshape = true); // 将数据进行反序列化,从磁盘导入之前存储的blob
void ToProto(BlobProto* proto, bool write_diff = false) const; // 将数据进行序列化,便于存储
protected:
shared_ptr<SyncedMemory> data_;
shared_ptr<SyncedMemory> diff_;
shared_ptr<SyncedMemory> shape_data_;
vector<int> shape_;
int count_;
int capacity_;
DISABLE_COPY_AND_ASSIGN(Blob);
}; // class Blob
回到Layer
Layer基类的Forward
方法,注意这并非是一个virtual方法,也就意味着它不希望子类对这个函数进行修改,即可以认为所有Layer都是使用的这个Forward函数,所以我们来看看具体的步骤:
template <typename Dtype>
class Layer {
public:
explicit Layer(const LayerParameter& param) : layer_param_(param) {
phase_ = param.phase();
if (layer_param_.blobs_size() > 0) {
blobs_.resize(layer_param_.blobs_size());
for (int i = 0; i < layer_param_.blobs_size(); ++i) {
blobs_[i].reset(new Blob<Dtype>());
blobs_[i]->FromProto(layer_param_.blobs(i));
}
}
}
virtual ~Layer() {}
void SetUp(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top) {
CheckBlobCounts(bottom, top);
LayerSetUp(bottom, top);
Reshape(bottom, top);
SetLossWeights(top);
}
/**
* @brief Does layer-specific setup: your layer should implement this function
* as well as Reshape.
*
* @param bottom
* the preshaped input blobs, whose data fields store the input data for
* this layer
* @param top
* the allocated but unshaped output blobs
*
* This method should do one-time layer specific setup. This includes reading
* and processing relevent parameters from the <code>layer_param_</code>.
* Setting up the shapes of top blobs and internal buffers should be done in
* <code>Reshape</code>, which will be called before the forward pass to
* adjust the top blob sizes.
*/
virtual void LayerSetUp(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top) {}
/**
* @brief Adjust the shapes of top blobs and internal buffers to accommodate
* the shapes of the bottom blobs.
*
* @param bottom the input blobs, with the requested input shapes
* @param top the top blobs, which should be reshaped as needed
*
* This method should reshape top blobs as needed according to the shapes
* of the bottom (input) blobs, as well as reshaping any internal buffers
* and making any other necessary adjustments so that the layer can
* accommodate the bottom blobs.
*/
virtual void Reshape(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top) = 0;
/**
* @brief Given the bottom blobs, compute the top blobs and the loss.
*
* @param bottom
* the input blobs, whose data fields store the input data for this layer
* @param top
* the preshaped output blobs, whose data fields will store this layers'
* outputs
*
eturn The total loss from the layer.
*
* The Forward wrapper calls the relevant device wrapper function
* (Forward_cpu or Forward_gpu) to compute the top blob values given the
* bottom blobs. If the layer has any non-zero loss_weights, the wrapper
* then computes and returns the loss.
*
* Your layer should implement Forward_cpu and (optionally) Forward_gpu.
*/
inline Dtype Forward(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top);
/**
* @brief Given the top blob error gradients, compute the bottom blob error
* gradients.
*
* @param top
* the output blobs, whose diff fields store the gradient of the error
* with respect to themselves
* @param propagate_down
* a vector with equal length to bottom, with each index indicating
* whether to propagate the error gradients down to the bottom blob at
* the corresponding index
* @param bottom
* the input blobs, whose diff fields will store the gradient of the error
* with respect to themselves after Backward is run
*
* The Backward wrapper calls the relevant device wrapper function
* (Backward_cpu or Backward_gpu) to compute the bottom blob diffs given the
* top blob diffs.
*
* Your layer should implement Backward_cpu and (optionally) Backward_gpu.
*/
inline void Backward(const vector<Blob<Dtype>*>& top,
const vector<bool>& propagate_down,
const vector<Blob<Dtype>*>& bottom);
vector<shared_ptr<Blob<Dtype> > >& blobs() {
return blobs_;
}
const LayerParameter& layer_param() const { return layer_param_; }
protected:
/** The protobuf that stores the layer parameters */
LayerParameter layer_param_; //层的参数: 卷积核大小,步长
Phase phase_;
/** The vector that stores the learnable parameters as a set of blobs. */
vector<shared_ptr<Blob<Dtype> > > blobs_; //滤波器参数
vector<bool> param_propagate_down_;
vector<Dtype> loss_;
virtual void Forward_cpu(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top) = 0;
virtual void Forward_gpu(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top) {
// LOG(WARNING) << "Using CPU code as backup.";
return Forward_cpu(bottom, top);
}
virtual void Backward_cpu(const vector<Blob<Dtype>*>& top,
const vector<bool>& propagate_down,
const vector<Blob<Dtype>*>& bottom) = 0;
virtual void Backward_gpu(const vector<Blob<Dtype>*>& top,
const vector<bool>& propagate_down,
const vector<Blob<Dtype>*>& bottom) {
// LOG(WARNING) << "Using CPU code as backup.";
Backward_cpu(top, propagate_down, bottom);
}
private:
DISABLE_COPY_AND_ASSIGN(Layer);
};
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:
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;
}
在Layer中比较重要的几个函数,Setup
, LayerSetup
, Reshape
, Forward
, BackWard
, Forward_cpu
, Forward_gpu
, Backward_cpu
, Backward_gpu
。
Reshape
,Forward_cpu
,Backward_cpu
函数是纯虚函数,子类一定要对其进行实现;LayerSetup
,Forward_gpu
,Backward_gpu
是虚函数,可以根据需要进行重写。Setup
,Forward
,BackWard
是普通函数,不要重写;
由于卷积也有许多种,所以在中间加了BaseConvolutionLayer
类,做为所有卷积类的基类。实现了如下函数,并将Reshape
函数由纯虚函数变为了虚函数:
LayerSetUp(const vector<Blob<Dtype>*>& bottom, const vector<Blob<Dtype>*>& top)
Reshape(const vector<Blob<Dtype>*>& bottom, const vector<Blob<Dtype>*>& top)
forward_cpu_gemm(const Dtype* input, const Dtype* weights, Dtype* output, bool skip_im2col)
forward_cpu_bias(Dtype* output, const Dtype* bias)
backward_cpu_gemm(const Dtype* output, const Dtype* weights, Dtype* input)
weight_cpu_gemm(const Dtype* input, const Dtype* output, Dtype* weights)
backward_cpu_bias(Dtype* bias, const Dtype* input)
forward_gpu_gemm(const Dtype* input, const Dtype* weights, Dtype* output, bool skip_im2col)
forward_gpu_bias(Dtype* output, const Dtype* bias)
backward_gpu_gemm(const Dtype* output, const Dtype* weights, Dtype* input)
weight_gpu_gemm(const Dtype* input, const Dtype* output, Dtype* weights)
backward_gpu_bias(Dtype* bias, const Dtype* input)
ConvolutionLayer
继承BaseConvolutionLayer
,实现了如下函数:
Forward_cpu(const vector<Blob<Dtype>*>& bottom, const vector<Blob<Dtype>*>& top)
Backward_cpu(const vector<Blob<Dtype>*>& top, const vector<bool>& propagate_down, const vector<Blob<Dtype>*>& bottom)
在Layer的Forward
函数中,首先调用Reshape
函数,这时调用的是BaseConvolutionLayer::Reshape
函数,caffe的数据组织类型为Blob,在输入(bottom)大小已知,卷积参数已知的情况下,是可以计算输出(top)的Blob的shape,如下:
// Shape the tops.
bottom_shape_ = &bottom[0]->shape();
compute_output_shape();
vector<int> top_shape(bottom[0]->shape().begin(),
bottom[0]->shape().begin() + channel_axis_);
top_shape.push_back(num_output_);
for (int i = 0; i < num_spatial_axes_; ++i) {
top_shape.push_back(output_shape_[i]);
}
for (int top_id = 0; top_id < top.size(); ++top_id) {
top[top_id]->Reshape(top_shape);
}
里面对每个输出top[i]调用了其成员函数Reshape,Blob的Reshape函数如下:
template <typename Dtype>
void Blob<Dtype>::Reshape(const vector<int>& shape) {
CHECK_LE(shape.size(), kMaxBlobAxes);
count_ = 1;
shape_.resize(shape.size());
if (!shape_data_ || shape_data_->size() < shape.size() * sizeof(int)) {
shape_data_.reset(new SyncedMemory(shape.size() * sizeof(int)));
}
int* shape_data = static_cast<int*>(shape_data_->mutable_cpu_data());
for (int i = 0; i < shape.size(); ++i) {
CHECK_GE(shape[i], 0);
if (count_ != 0) {
CHECK_LE(shape[i], INT_MAX / count_) << "blob size exceeds INT_MAX";
}
count_ *= shape[i];
shape_[i] = shape[i]; //拷到内部
shape_data[i] = shape[i];
}
if (count_ > capacity_) { //内存不够
capacity_ = count_;
data_.reset(new SyncedMemory(capacity_ * sizeof(Dtype))); //重新申请
diff_.reset(new SyncedMemory(capacity_ * sizeof(Dtype)));
}
}
其实就是将传入的shape
复制到Blob的内部变量shape_
中,并判断内存是否满足要求,不满足要求的话重新申请内存。
前向传播这里我们分析cpu的情况,Reshape
之后是Forward_cpu
,现在调用的是ConvolutionLayer::Forward_cpu
函数:
template <typename Dtype>
void ConvolutionLayer<Dtype>::Forward_cpu(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top) {
const Dtype* weight = this->blobs_[0]->cpu_data();
for (int i = 0; i < bottom.size(); ++i) {
const Dtype* bottom_data = bottom[i]->cpu_data();
Dtype* top_data = top[i]->mutable_cpu_data();
for (int n = 0; n < this->num_; ++n) {
this->forward_cpu_gemm(bottom_data + n * this->bottom_dim_, weight,
top_data + n * this->top_dim_);
if (this->bias_term_) {
const Dtype* bias = this->blobs_[1]->cpu_data();
this->forward_cpu_bias(top_data + n * this->top_dim_, bias);
}
}
}
}
代码主要是对于每个bottom、top,要做num_(batch_size)次矩阵乘法(forward_cpu_gemm
),将bottom_data与weight相乘,结果保存到top_data中,这里mutable_cpu_data
表示要对这个地址进行写数据,具体地矩阵乘法:
template <typename Dtype>
void BaseConvolutionLayer<Dtype>::forward_cpu_gemm(const Dtype* input,
const Dtype* weights, Dtype* output, bool skip_im2col) {
const Dtype* col_buff = input;
if (!is_1x1_) {
if (!skip_im2col) {
conv_im2col_cpu(input, col_buffer_.mutable_cpu_data());
}
col_buff = col_buffer_.cpu_data();
}
for (int g = 0; g < group_; ++g) {
caffe_cpu_gemm<Dtype>(CblasNoTrans, CblasNoTrans, conv_out_channels_ /
group_, conv_out_spatial_dim_, kernel_dim_,
(Dtype)1., weights + weight_offset_ * g, col_buff + col_offset_ * g,
(Dtype)0., output + output_offset_ * g);
}
}
这里有conv_im2col_cpu
函数。如果我们不进行转换,我们需要循环进行多次矩阵乘法,这里使用这个函数将每个patch(kxkxC)拉直,然后将这些patch堆在一起,这样就可以只进行一次卷积就可以求出所有结果,caffe_cpu_gemm就是封装的cblas的矩阵乘法ouput = weights * col_buff
。
再回到Forward
函数中,做完Forward_cpu
后,会遍历所有层判断是否是loss层,如果是则根据cpu_diff()
计算loss:
inline Dtype Layer<Dtype>::Forward(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top) {
Dtype loss = 0;
Reshape(bottom, top);
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);
}
}
这样Forward
函数就结束了,下面开始进入Backward
函数,直接来看Layer
的Backward
函数,如下:
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.";
}
}
里面直接调用Backward_cpu
函数,来看ConvolutionLayer
的Backward_cpu
函数,如下:
template <typename Dtype>
void ConvolutionLayer<Dtype>::Backward_cpu(const vector<Blob<Dtype>*>& top,
const vector<bool>& propagate_down, const vector<Blob<Dtype>*>& bottom) {
const Dtype* weight = this->blobs_[0]->cpu_data();
Dtype* weight_diff = this->blobs_[0]->mutable_cpu_diff();
for (int i = 0; i < top.size(); ++i) {
const Dtype* top_diff = top[i]->cpu_diff();
const Dtype* bottom_data = bottom[i]->cpu_data();
Dtype* bottom_diff = bottom[i]->mutable_cpu_diff();
// Bias gradient, if necessary.
if (this->bias_term_ && this->param_propagate_down_[1]) {
Dtype* bias_diff = this->blobs_[1]->mutable_cpu_diff();
for (int n = 0; n < this->num_; ++n) {
this->backward_cpu_bias(bias_diff, top_diff + n * this->top_dim_);
}
}
if (this->param_propagate_down_[0] || propagate_down[i]) {
for (int n = 0; n < this->num_; ++n) {
// gradient w.r.t. weight. Note that we will accumulate diffs.
if (this->param_propagate_down_[0]) {
this->weight_cpu_gemm(bottom_data + n * this->bottom_dim_,
top_diff + n * this->top_dim_, weight_diff);
}
// gradient w.r.t. bottom data, if necessary.
if (propagate_down[i]) {
this->backward_cpu_gemm(top_diff + n * this->top_dim_, weight,
bottom_diff + n * this->bottom_dim_);
}
}
}
}
}
里面根据top_diff
分别更新了当前层的weight_diff
(weight_cpu_gemm
),和bottom_diff
(backward_cpu_gemm
)(计算bottom_diff实际上是为了weight_diff)。
那么Backward也结束了,它分别计算了各层的权重参数的梯度(weight_diff)、以及各层blob的梯度(bottom_diff)。
再回到solver.Solver
函数中,发现下面是执行ApplyUpdate()
函数,才是真正更新参数的时候,solver.ApplyUpdate()
实际上调用了Net.Update()
函数,如下:
template <typename Dtype>
void Net<Dtype>::Update() {
for (int i = 0; i < learnable_params_.size(); ++i) {
learnable_params_[i]->Update();
}
}
这里的learnable_params_
实际上就是每层可训练的参数,也就是每层的权重参数Blob,我们之前更新了这些Blob里的diff
值,那我们再继续看看Blob.Update()
函数里做了什么:
void Blob<Dtype>::Update() {
// We will perform update based on where the data is located.
switch (data_->head()) {
case SyncedMemory::HEAD_AT_CPU:
// perform computation on CPU
caffe_axpy<Dtype>(count_, Dtype(-1),
static_cast<const Dtype*>(diff_->cpu_data()),
static_cast<Dtype*>(data_->mutable_cpu_data()));
break;
//...
}
}
主要是做了如下的计算data_ = data_ - diff_
,caffe_axpy
实际上是封装了cblas的函数,主要做两个函数相加,由于传入的系数是Dtype(-1)
,所以是进行了相减更新data_
,至此,每层的权重参数都得到了更新,那么一次迭代更新也就结束了。下面就是多次调用这个过程,直到训练得到一个较好的权重参数。
test阶段
测试test阶段,不需要solver,直接使用Net进行Forward就可以得到结果:
Net<float> caffe_net(FLAGS_model, caffe::TEST, FLAGS_level, &stages);
const vector<Blob<float>*>& result = caffe_net.Forward(&iter_loss);
pycaffe
首先有一个_caffe.cpp文件,里面将所有caffe框架编译成一个_caffe.so,而pycaffe.py相当于一个wrapper,封装了一些python接口。pycaffe中可以将_caffe.so
中的对象import进来,当作python对象使用,如下:
from ._caffe import Net, SGDSolver, NesterovSolver, AdaGradSolver,
RMSPropSolver, AdaDeltaSolver, AdamSolver, NCCL, Timer
之所以可以导入直接使用,这是因为在_caffe.cpp中使用BOOST_PYTHON_MODULE
进行了导出:
BOOST_PYTHON_MODULE(_caffe) {
...
}
如下是导出一个类的方法:
#include<string>
#include<boost/python.hpp>
using namespace std;
using namespace boost::python;
struct World
{
void set(string msg) { this->msg = msg; }
string greet() { return msg; }
string msg;
};
BOOST_PYTHON_MODULE(hello) //导出的module 名字
{
class_<World>("World")
.def("greet", &World::greet)
.def("set", &World::set);
}
如下是python中调用导出的方法:
import hello
planet = hello.World() # 调用默认构造函数,产生类对象
planet.set("howdy") # 调用对象的方法
print planet.greet() # 调用对象的方法
如果不想导出任何构造函数,则使用no_init:
class_<Abstract>("Abstract",no_init)
最后,caffe目录中提供了一个__init__.py
文件,将整个caffe目录变成一个python包:
from .pycaffe import Net, SGDSolver, NesterovSolver, AdaGradSolver, RMSPropSolver, AdaDeltaSolver, AdamSolver, NCCL, Timer
from ._caffe import init_log, log, set_mode_cpu, set_mode_gpu, set_device, Layer, get_solver, layer_type_list, set_random_seed, solver_count, set_solver_count, solver_rank, set_solver_rank, set_multiprocess, has_nccl
from ._caffe import __version__
from .proto.caffe_pb2 import TRAIN, TEST
from .classifier import Classifier
from .detector import Detector
from . import io
from .net_spec import layers, params, NetSpec, to_proto
这样,外面就可以使用caffe.Net, caffe.init_log, caffe.__version__, caffe.TRAIN, caffe.Classifier caffe.Detector caffe.io...
去使用caffe的Python接口了。