在了解了blob,layer.hpp,layer.cpp和im2col后，贴上base_conv.hpp和.cpp。
#ifndef CAFFE_BASE_CONVOLUTION_LAYER_HPP_
#define CAFFE_BASE_CONVOLUTION_LAYER_HPP_
#include <vector>
#include "caffe/blob.hpp"
#include "caffe/layer.hpp"
#include "caffe/proto/caffe.pb.h"
#include "caffe/util/im2col.hpp"
namespace caffe {
/**
* @brief Abstract base class that factors out the BLAS code common to
* ConvolutionLayer and DeconvolutionLayer.
*/
template <typename Dtype>
class BaseConvolutionLayer : public Layer<Dtype> {
public:
explicit BaseConvolutionLayer(const LayerParameter& param)
: Layer<Dtype>(param) {}
virtual void LayerSetUp(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top);
virtual void Reshape(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top);
virtual inline int MinBottomBlobs() const { return 1; }
virtual inline int MinTopBlobs() const { return 1; }
virtual inline bool EqualNumBottomTopBlobs() const { return true; }// 这三个函数为了检验bottom，top blob的个数
protected:
// Helper functions that abstract away the column buffer and gemm arguments.
// The last argument in forward_cpu_gemm is so that we can skip the im2col if
// we just called weight_cpu_gemm with the same input.
void forward_cpu_gemm(const Dtype* input, const Dtype* weights,
Dtype* output, bool skip_im2col = false);
void forward_cpu_bias(Dtype* output, const Dtype* bias);
void backward_cpu_gemm(const Dtype* input, const Dtype* weights,
Dtype* output);
void weight_cpu_gemm(const Dtype* input, const Dtype* output, Dtype*
weights);
void backward_cpu_bias(Dtype* bias, const Dtype* input);
#ifndef CPU_ONLY
void forward_gpu_gemm(const Dtype* col_input, const Dtype* weights,
Dtype* output, bool skip_im2col = false);
void forward_gpu_bias(Dtype* output, const Dtype* bias);
void backward_gpu_gemm(const Dtype* input, const Dtype* weights,
Dtype* col_output);
void weight_gpu_gemm(const Dtype* col_input, const Dtype* output, Dtype*
weights);
void backward_gpu_bias(Dtype* bias, const Dtype* input);
#endif
/// @brief The spatial dimensions of the input.
inline int input_shape(int i) {
return (*bottom_shape_)[channel_axis_ + i]; //索引是从channel_axis+1开始，所以是为了得到bottom的height和width
}
// reverse_dimensions should return true iff we are implementing deconv, so
// that conv helpers know which dimensions are which.
virtual bool reverse_dimensions() = 0; //反卷积标志位，在conv_layer.hpp直接置False
// Compute height_out_ and width_out_ from other parameters.
virtual void compute_output_shape() = 0;
/// @brief The spatial dimensions of a filter kernel.
Blob<int> kernel_shape_; // spatial dimensions,所以只有height和width
/// @brief The spatial dimensions of the stride.
Blob<int> stride_; // [stride_h,stride_w]
/// @brief The spatial dimensions of the padding.
Blob<int> pad_; //[pad_h,pad_w]
/// @brief The spatial dimensions of the dilation.
Blob<int> dilation_; //空洞卷积系数 _h,_w
/// @brief The spatial dimensions of the convolution input.
Blob<int> conv_input_shape_; // [input_channels,inp_height,inp_width]
/// @brief The spatial dimensions of the col_buffer.
vector<int> col_buffer_shape_; // im2col后col矩阵的形状，[input_channels*kernel_h*kernel_w, output_h*output_w]
/// @brief The spatial dimensions of the output.
vector<int> output_shape_; //top blob 的shape
const vector<int>* bottom_shape_; // bottom blob的shape,[batch,channel,h,w]
int num_spatial_axes_; //spatial_axes，所以一般为2，在随笔记中已经解释
int bottom_dim_; //bottom的count/batch
int top_dim_; //top的count/batch
int channel_axis_; //channel轴的序号,一般是1，[N,C,H,W]
int num_; // batchsizes
int channels_; // input_channels
int group_;
int out_spatial_dim_; //output_h*output_w
int weight_offset_;//group时用到的参数
int num_output_;
bool bias_term_;
bool is_1x1_;
bool force_nd_im2col_; // 强制n维卷积，caffe.proto默认值是false
private:
// wrap im2col/col2im so we don't have to remember the (long) argument lists
inline void conv_im2col_cpu(const Dtype* data, Dtype* col_buff) {
if (!force_nd_im2col_ && num_spatial_axes_ == 2) {
im2col_cpu(data, conv_in_channels_, //conv_in_channels_ = channels
conv_input_shape_.cpu_data()[1], conv_input_shape_.cpu_data()[2],
kernel_shape_.cpu_data()[0], kernel_shape_.cpu_data()[1],
pad_.cpu_data()[0], pad_.cpu_data()[1],
stride_.cpu_data()[0], stride_.cpu_data()[1],
dilation_.cpu_data()[0], dilation_.cpu_data()[1], col_buff);
} else {
im2col_nd_cpu(data, num_spatial_axes_, conv_input_shape_.cpu_data(),
col_buffer_shape_.data(), kernel_shape_.cpu_data(),
pad_.cpu_data(), stride_.cpu_data(), dilation_.cpu_data(), col_buff);
}
} //CV一般用的都是二维卷积，随笔记中已经解释了，所以一般都是调用im2col_cpu函数，得到col矩阵
inline void conv_col2im_cpu(const Dtype* col_buff, Dtype* data) {
if (!force_nd_im2col_ && num_spatial_axes_ == 2) {
col2im_cpu(col_buff, conv_in_channels_,
conv_input_shape_.cpu_data()[1], conv_input_shape_.cpu_data()[2],
kernel_shape_.cpu_data()[0], kernel_shape_.cpu_data()[1],
pad_.cpu_data()[0], pad_.cpu_data()[1],
stride_.cpu_data()[0], stride_.cpu_data()[1],
dilation_.cpu_data()[0], dilation_.cpu_data()[1], data);
} else {
col2im_nd_cpu(col_buff, num_spatial_axes_, conv_input_shape_.cpu_data(),
col_buffer_shape_.data(), kernel_shape_.cpu_data(),
pad_.cpu_data(), stride_.cpu_data(), dilation_.cpu_data(), data);
}
}
#ifndef CPU_ONLY
inline void conv_im2col_gpu(const Dtype* data, Dtype* col_buff) {
if (!force_nd_im2col_ && num_spatial_axes_ == 2) {
im2col_gpu(data, conv_in_channels_,
conv_input_shape_.cpu_data()[1], conv_input_shape_.cpu_data()[2],
kernel_shape_.cpu_data()[0], kernel_shape_.cpu_data()[1],
pad_.cpu_data()[0], pad_.cpu_data()[1],
stride_.cpu_data()[0], stride_.cpu_data()[1],
dilation_.cpu_data()[0], dilation_.cpu_data()[1], col_buff);
} else {
im2col_nd_gpu(data, num_spatial_axes_, num_kernels_im2col_,
conv_input_shape_.gpu_data(), col_buffer_.gpu_shape(),
kernel_shape_.gpu_data(), pad_.gpu_data(),
stride_.gpu_data(), dilation_.gpu_data(), col_buff);
}
}
inline void conv_col2im_gpu(const Dtype* col_buff, Dtype* data) {
if (!force_nd_im2col_ && num_spatial_axes_ == 2) {
col2im_gpu(col_buff, conv_in_channels_,
conv_input_shape_.cpu_data()[1], conv_input_shape_.cpu_data()[2],
kernel_shape_.cpu_data()[0], kernel_shape_.cpu_data()[1],
pad_.cpu_data()[0], pad_.cpu_data()[1],
stride_.cpu_data()[0], stride_.cpu_data()[1],
dilation_.cpu_data()[0], dilation_.cpu_data()[1], data);
} else {
col2im_nd_gpu(col_buff, num_spatial_axes_, num_kernels_col2im_,
conv_input_shape_.gpu_data(), col_buffer_.gpu_shape(),
kernel_shape_.gpu_data(), pad_.gpu_data(), stride_.gpu_data(),
dilation_.gpu_data(), data);
}
}
#endif
int num_kernels_im2col_; //
int num_kernels_col2im_;//这两个参数对应n维GPU实现，cpp文件有解释，不用太在意
int conv_out_channels_; // output参数，prototxt中设置
int conv_in_channels_; //输入channel
int conv_out_spatial_dim_;// output_h*output_w
int kernel_dim_; //一个三维卷积核的大小
int col_offset_;
int output_offset_;//group时用到的两个参数
Blob<Dtype> col_buffer_;
Blob<Dtype> bias_multiplier_; //偏置扩展为矩阵
};
} // namespace caffe
#endif // CAFFE_BASE_CONVOLUTION_LAYER_HPP_ 接下来是cpp
#include <algorithm>
#include <vector>
#include "caffe/filler.hpp"
#include "caffe/layers/base_conv_layer.hpp"
#include "caffe/util/im2col.hpp"
#include "caffe/util/math_functions.hpp"
namespace caffe {
template <typename Dtype>
void BaseConvolutionLayer<Dtype>::LayerSetUp(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top) {
// Configure the kernel size, padding, stride, and inputs.
ConvolutionParameter conv_param = this->layer_param_.convolution_param();
force_nd_im2col_ = conv_param.force_nd_im2col();//默认是False
//CanonicalAxisIndex函数的作用是索引标准化，即把[-N,N)转换为[0,N),axis在proto中默认为1，故channel_axis_=axis_默认为1
channel_axis_ = bottom[0]->CanonicalAxisIndex(conv_param.axis());
const int first_spatial_axis = channel_axis_ + 1;//图像第一个spatial_axis，1+1=2，一般是h
const int num_axes = bottom[0]->num_axes();
num_spatial_axes_ = num_axes - first_spatial_axis;//4-2，所以hpp中说num_spatial_axes_一般为2
CHECK_GE(num_spatial_axes_, 0);//保证图像的spatial_axes大于0,其实spatial_axes就是卷积操作的最小维度，h和w
vector<int> bottom_dim_blob_shape(1, num_spatial_axes_ + 1); //卷积操作输入维度3维
vector<int> spatial_dim_blob_shape(1, std::max(num_spatial_axes_, 1));
// Setup filter kernel dimensions (kernel_shape_).
kernel_shape_.Reshape(spatial_dim_blob_shape);//这两行用来初始化卷积核的形状，h和w
int* kernel_shape_data = kernel_shape_.mutable_cpu_data();
//根据kernel_h，kernel_w或者kernel_size来赋值kernel_shape_data，但是不能混用
if (conv_param.has_kernel_h() || conv_param.has_kernel_w()) {
CHECK_EQ(num_spatial_axes_, 2)
<< "kernel_h & kernel_w can only be used for 2D convolution.";
CHECK_EQ(0, conv_param.kernel_size_size())
<< "Either kernel_size or kernel_h/w should be specified; not both.";
kernel_shape_data[0] = conv_param.kernel_h();
kernel_shape_data[1] = conv_param.kernel_w();
} else {
const int num_kernel_dims = conv_param.kernel_size_size();
CHECK(num_kernel_dims == 1 || num_kernel_dims == num_spatial_axes_)
<< "kernel_size must be specified once, or once per spatial dimension "
<< "(kernel_size specified " << num_kernel_dims << " times; "
<< num_spatial_axes_ << " spatial dims).";
for (int i = 0; i < num_spatial_axes_; ++i) {
kernel_shape_data[i] =
conv_param.kernel_size((num_kernel_dims == 1) ? 0 : i);
}
}
for (int i = 0; i < num_spatial_axes_; ++i) {
CHECK_GT(kernel_shape_data[i], 0) << "Filter dimensions must be nonzero.";
}
// Setup stride dimensions (stride_).
stride_.Reshape(spatial_dim_blob_shape);
int* stride_data = stride_.mutable_cpu_data();
//stride_data赋值，和kernel_shape_data类似，若prototxt中没有stride参数，默认为[1,1]
if (conv_param.has_stride_h() || conv_param.has_stride_w()) {
CHECK_EQ(num_spatial_axes_, 2)
<< "stride_h & stride_w can only be used for 2D convolution.";
CHECK_EQ(0, conv_param.stride_size())
<< "Either stride or stride_h/w should be specified; not both.";
stride_data[0] = conv_param.stride_h();
stride_data[1] = conv_param.stride_w();
} else {
const int num_stride_dims = conv_param.stride_size();
CHECK(num_stride_dims == 0 || num_stride_dims == 1 ||
num_stride_dims == num_spatial_axes_)
<< "stride must be specified once, or once per spatial dimension "
<< "(stride specified " << num_stride_dims << " times; "
<< num_spatial_axes_ << " spatial dims).";
const int kDefaultStride = 1;
for (int i = 0; i < num_spatial_axes_; ++i) {
stride_data[i] = (num_stride_dims == 0) ? kDefaultStride :
conv_param.stride((num_stride_dims == 1) ? 0 : i);
CHECK_GT(stride_data[i], 0) << "Stride dimensions must be nonzero.";
}
}
// Setup pad dimensions (pad_).
//pad_data赋值，和stride_data类似，prototxt中没有pad参数，默认为[0,0]
pad_.Reshape(spatial_dim_blob_shape);
int* pad_data = pad_.mutable_cpu_data();
if (conv_param.has_pad_h() || conv_param.has_pad_w()) {
CHECK_EQ(num_spatial_axes_, 2)
<< "pad_h & pad_w can only be used for 2D convolution.";
CHECK_EQ(0, conv_param.pad_size())
<< "Either pad or pad_h/w should be specified; not both.";
pad_data[0] = conv_param.pad_h();
pad_data[1] = conv_param.pad_w();
} else {
const int num_pad_dims = conv_param.pad_size();
CHECK(num_pad_dims == 0 || num_pad_dims == 1 ||
num_pad_dims == num_spatial_axes_)
<< "pad must be specified once, or once per spatial dimension "
<< "(pad specified " << num_pad_dims << " times; "
<< num_spatial_axes_ << " spatial dims).";
const int kDefaultPad = 0;
for (int i = 0; i < num_spatial_axes_; ++i) {
pad_data[i] = (num_pad_dims == 0) ? kDefaultPad :
conv_param.pad((num_pad_dims == 1) ? 0 : i);
}
}
// Setup dilation dimensions (dilation_).
//同pad和stride，缺省时默认[1,1]，表示卷积核不扩展
dilation_.Reshape(spatial_dim_blob_shape);
int* dilation_data = dilation_.mutable_cpu_data();
const int num_dilation_dims = conv_param.dilation_size();
CHECK(num_dilation_dims == 0 || num_dilation_dims == 1 ||
num_dilation_dims == num_spatial_axes_)
<< "dilation must be specified once, or once per spatial dimension "
<< "(dilation specified " << num_dilation_dims << " times; "
<< num_spatial_axes_ << " spatial dims).";
const int kDefaultDilation = 1;
for (int i = 0; i < num_spatial_axes_; ++i) {
dilation_data[i] = (num_dilation_dims == 0) ? kDefaultDilation :
conv_param.dilation((num_dilation_dims == 1) ? 0 : i);
}
// Special case: im2col is the identity for 1x1 convolution with stride 1
// and no padding, so flag for skipping the buffer and transformation.
//判断 1*1卷积
is_1x1_ = true;
for (int i = 0; i < num_spatial_axes_; ++i) {
is_1x1_ &=
kernel_shape_data[i] == 1 && stride_data[i] == 1 && pad_data[i] == 0;
if (!is_1x1_) { break; }
}
// Configure output channels and groups.
channels_ = bottom[0]->shape(channel_axis_);
num_output_ = this->layer_param_.convolution_param().num_output();
CHECK_GT(num_output_, 0);
group_ = this->layer_param_.convolution_param().group();
CHECK_EQ(channels_ % group_, 0);
CHECK_EQ(num_output_ % group_, 0)
<< "Number of output should be multiples of group.";//每个group的输出channel,和输出channel大于0
//反转卷积的话，交换输入、输出的channels
if (reverse_dimensions()) {
conv_out_channels_ = channels_;
conv_in_channels_ = num_output_;
} else {
conv_out_channels_ = num_output_;
conv_in_channels_ = channels_;
}
// Handle the parameters: weights and biases.
// - blobs_[0] holds the filter weights
// - blobs_[1] holds the biases (optional)
vector<int> weight_shape(2);
weight_shape[0] = conv_out_channels_; //卷积核的个数只和输出channels相关，所以不用除group
weight_shape[1] = conv_in_channels_ / group_;
for (int i = 0; i < num_spatial_axes_; ++i) {
weight_shape.push_back(kernel_shape_data[i]);
} //weight_shape现在有4个元素,[conv_out_channels_,conv_in_channels_ / group_,k_h,k_w]
bias_term_ = this->layer_param_.convolution_param().bias_term();
vector<int> bias_shape(bias_term_, num_output_);//true相当于1，,bias[0]=num_output_
//接下来的if else完成了对权值的处理，blobs_存放w,b，如果是从头训练，blobs_.size为0，调用else
//的代码段，完成对blobs的初始化和权重填充；若w,b已经有参数了，执行if中的几个if判断w，b顺序等函数
if (this->blobs_.size() > 0) {
CHECK_EQ(1 + bias_term_, this->blobs_.size())
<< "Incorrect number of weight blobs.";//检查conv层参数个数是不是1+b,也就是检测w是不是1个
//检查w是不是blobs[0]，如果有的话
if (weight_shape != this->blobs_[0]->shape()) {
Blob<Dtype> weight_shaped_blob(weight_shape);
LOG(FATAL) << "Incorrect weight shape: expected shape "
<< weight_shaped_blob.shape_string() << "; instead, shape was "
<< this->blobs_[0]->shape_string();
}
//检查b是不是blobs[1],如果有的话
if (bias_term_ && bias_shape != this->blobs_[1]->shape()) {
Blob<Dtype> bias_shaped_blob(bias_shape);
LOG(FATAL) << "Incorrect bias shape: expected shape "
<< bias_shaped_blob.shape_string() << "; instead, shape was "
<< this->blobs_[1]->shape_string();
}
LOG(INFO) << "Skipping parameter initialization";
} else {
if (bias_term_) {
this->blobs_.resize(2);
} else {
this->blobs_.resize(1);
}
// Initialize and fill the weights:
// output channels x input channels per-group x kernel height x kernel width
this->blobs_[0].reset(new Blob<Dtype>(weight_shape));//将blobs_[0]大小初始化为weight_shape
shared_ptr<Filler<Dtype> > weight_filler(GetFiller<Dtype>(
this->layer_param_.convolution_param().weight_filler()));//读取我们定义层的参数中的权重填充，默认为0
weight_filler->Fill(this->blobs_[0].get());//权重填充
// If necessary, initialize and fill the biases.
if (bias_term_) { //bias填充
this->blobs_[1].reset(new Blob<Dtype>(bias_shape));
shared_ptr<Filler<Dtype> > bias_filler(GetFiller<Dtype>(
this->layer_param_.convolution_param().bias_filler()));
bias_filler->Fill(this->blobs_[1].get());
}
}
kernel_dim_ = this->blobs_[0]->count(1);//相当于count(1,4),channels_/group_ *k_h*k_w
weight_offset_ = conv_out_channels_ * kernel_dim_ / group_;
// Propagate gradients to the parameters (as directed by backward pass).
this->param_propagate_down_.resize(this->blobs_.size(), true);//初始化时，一般都会打开梯度反传的开关
}
template <typename Dtype>
void BaseConvolutionLayer<Dtype>::Reshape(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top) {
const int first_spatial_axis = channel_axis_ + 1;//1+1
CHECK_EQ(bottom[0]->num_axes(), first_spatial_axis + num_spatial_axes_)
<< "bottom num_axes may not change.";
num_ = bottom[0]->count(0, channel_axis_);//batchsizes
CHECK_EQ(bottom[0]->shape(channel_axis_), channels_)
<< "Input size incompatible with convolution kernel.";
// TODO: generalize to handle inputs of different shapes.
//所有输入blob有相同的size
for (int bottom_id = 1; bottom_id < bottom.size(); ++bottom_id) {
CHECK(bottom[0]->shape() == bottom[bottom_id]->shape())
<< "All inputs must have the same shape.";
}
// 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_);//第一个元素是batchsize
top_shape.push_back(num_output_);
for (int i = 0; i < num_spatial_axes_; ++i) {
top_shape.push_back(output_shape_[i]);
} //top shape 4个元素均写好
for (int top_id = 0; top_id < top.size(); ++top_id) {
top[top_id]->Reshape(top_shape);
}
if (reverse_dimensions()) {
conv_out_spatial_dim_ = bottom[0]->count(first_spatial_axis);
} else {
conv_out_spatial_dim_ = top[0]->count(first_spatial_axis);//output_h*w
}
col_offset_ = kernel_dim_ * conv_out_spatial_dim_;
output_offset_ = conv_out_channels_ * conv_out_spatial_dim_ / group_;
// Setup input dimensions (conv_input_shape_).
vector<int> bottom_dim_blob_shape(1, num_spatial_axes_ + 1);
conv_input_shape_.Reshape(bottom_dim_blob_shape);
int* conv_input_shape_data = conv_input_shape_.mutable_cpu_data();
for (int i = 0; i < num_spatial_axes_ + 1; ++i) {
if (reverse_dimensions()) {
conv_input_shape_data[i] = top[0]->shape(channel_axis_ + i);
} else
conv_input_shape_data[i] = bottom[0]->shape(channel_axis_ + i);// conv_input_shape_data存放bottom blob.shape[1,2,3]
}
}
// The im2col result buffer will only hold one image at a time to avoid
// overly large memory usage. In the special case of 1x1 convolution
// it goes lazily unused to save memory.
col_buffer_shape_.clear();
col_buffer_shape_.push_back(kernel_dim_ * group_);//input_channels*k_h*k_w
for (int i = 0; i < num_spatial_axes_; ++i) {
if (reverse_dimensions()) {
col_buffer_shape_.push_back(input_shape(i + 1));
} else {
col_buffer_shape_.push_back(output_shape_[i]);
}
}//col_buffer3维，生成的矩阵行数为kernel_dim_ * group_，列数为后两个元素的乘积，output_h*w
col_buffer_.Reshape(col_buffer_shape_);
bottom_dim_ = bottom[0]->count(channel_axis_);
top_dim_ = top[0]->count(channel_axis_);
num_kernels_im2col_ = conv_in_channels_ * conv_out_spatial_dim_;
num_kernels_col2im_ = reverse_dimensions() ? top_dim_ : bottom_dim_;
// Set up the all ones "bias multiplier" for adding biases by BLAS
out_spatial_dim_ = top[0]->count(first_spatial_axis);
if (bias_term_) {
vector<int> bias_multiplier_shape(1, out_spatial_dim_);
bias_multiplier_.Reshape(bias_multiplier_shape);
caffe_set(bias_multiplier_.count(), Dtype(1),
bias_multiplier_.mutable_cpu_data());
}
}
//卷积层的前向传播，CPU
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);
}
}
//前向传播加入bias，CPU
template <typename Dtype>
void BaseConvolutionLayer<Dtype>::forward_cpu_bias(Dtype* output,
const Dtype* bias) {
caffe_cpu_gemm<Dtype>(CblasNoTrans, CblasNoTrans, num_output_,
out_spatial_dim_, 1, (Dtype)1., bias, bias_multiplier_.cpu_data(),
(Dtype)1., output);
}
//卷积层的反向传播，CPU
template <typename Dtype>
void BaseConvolutionLayer<Dtype>::backward_cpu_gemm(const Dtype* output,
const Dtype* weights, Dtype* input) {
Dtype* col_buff = col_buffer_.mutable_cpu_data();
if (is_1x1_) {
col_buff = input;
}
for (int g = 0; g < group_; ++g) {
caffe_cpu_gemm<Dtype>(CblasTrans, CblasNoTrans, kernel_dim_,
conv_out_spatial_dim_, conv_out_channels_ / group_,
(Dtype)1., weights + weight_offset_ * g, output + output_offset_ * g,
(Dtype)0., col_buff + col_offset_ * g);
}
if (!is_1x1_) {
conv_col2im_cpu(col_buff, input);//im2col反向操作
}
}
//计算权重的增量，反向传播的时候用到，CPU
template <typename Dtype>
void BaseConvolutionLayer<Dtype>::weight_cpu_gemm(const Dtype* input,
const Dtype* output, Dtype* weights) {
const Dtype* col_buff = input;
if (!is_1x1_) {
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, CblasTrans, conv_out_channels_ / group_,
kernel_dim_, conv_out_spatial_dim_,
(Dtype)1., output + output_offset_ * g, col_buff + col_offset_ * g,
(Dtype)1., weights + weight_offset_ * g);
}
}
template <typename Dtype>
void BaseConvolutionLayer<Dtype>::backward_cpu_bias(Dtype* bias,
const Dtype* input) {
caffe_cpu_gemv<Dtype>(CblasNoTrans, num_output_, out_spatial_dim_, 1.,
input, bias_multiplier_.cpu_data(), 1., bias);
}
#ifndef CPU_ONLY
template <typename Dtype>
void BaseConvolutionLayer<Dtype>::forward_gpu_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_gpu(input, col_buffer_.mutable_gpu_data());
}
col_buff = col_buffer_.gpu_data();
}
for (int g = 0; g < group_; ++g) {
caffe_gpu_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);
}
}
template <typename Dtype>
void BaseConvolutionLayer<Dtype>::forward_gpu_bias(Dtype* output,
const Dtype* bias) {
caffe_gpu_gemm<Dtype>(CblasNoTrans, CblasNoTrans, num_output_,
out_spatial_dim_, 1, (Dtype)1., bias, bias_multiplier_.gpu_data(),
(Dtype)1., output);
}
template <typename Dtype>
void BaseConvolutionLayer<Dtype>::backward_gpu_gemm(const Dtype* output,
const Dtype* weights, Dtype* input) {
Dtype* col_buff = col_buffer_.mutable_gpu_data();
if (is_1x1_) {
col_buff = input;
}
for (int g = 0; g < group_; ++g) {
caffe_gpu_gemm<Dtype>(CblasTrans, CblasNoTrans, kernel_dim_,
conv_out_spatial_dim_, conv_out_channels_ / group_,
(Dtype)1., weights + weight_offset_ * g, output + output_offset_ * g,
(Dtype)0., col_buff + col_offset_ * g);
}
if (!is_1x1_) {
conv_col2im_gpu(col_buff, input);
}
}
template <typename Dtype>
void BaseConvolutionLayer<Dtype>::weight_gpu_gemm(const Dtype* input,
const Dtype* output, Dtype* weights) {
const Dtype* col_buff = input;
if (!is_1x1_) {
conv_im2col_gpu(input, col_buffer_.mutable_gpu_data());
col_buff = col_buffer_.gpu_data();
}
for (int g = 0; g < group_; ++g) {
caffe_gpu_gemm<Dtype>(CblasNoTrans, CblasTrans, conv_out_channels_ / group_,
kernel_dim_, conv_out_spatial_dim_,
(Dtype)1., output + output_offset_ * g, col_buff + col_offset_ * g,
(Dtype)1., weights + weight_offset_ * g);
}
}
template <typename Dtype>
void BaseConvolutionLayer<Dtype>::backward_gpu_bias(Dtype* bias,
const Dtype* input) {
caffe_gpu_gemv<Dtype>(CblasNoTrans, num_output_, out_spatial_dim_, 1.,
input, bias_multiplier_.gpu_data(), 1., bias);
}
#endif // !CPU_ONLY
INSTANTIATE_CLASS(BaseConvolutionLayer);
} // namespace caffe