1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
|
#ifdef USE_CUDNN
#include <algorithm>
#include <vector>
#include "caffe/layer.hpp"
#include "caffe/vision_layers.hpp"
namespace caffe {
template <typename Dtype>
void CuDNNReLULayer<Dtype>::Forward_gpu(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top) {
// Fallback to standard Caffe for leaky ReLU.
if (ReLULayer<Dtype>::layer_param_.relu_param().negative_slope() != 0) {
return ReLULayer<Dtype>::Forward_gpu(bottom, top);
}
const Dtype* bottom_data = bottom[0]->gpu_data();
Dtype* top_data = top[0]->mutable_gpu_data();
Dtype alpha = 1.0;
Dtype beta = 0.0;
CUDNN_CHECK(cudnnActivationForward(this->handle_,
CUDNN_ACTIVATION_RELU,
reinterpret_cast<void *>(&alpha),
this->bottom_desc_, bottom_data,
reinterpret_cast<void *>(&beta),
this->top_desc_, top_data));
}
template <typename Dtype>
void CuDNNReLULayer<Dtype>::Backward_gpu(const vector<Blob<Dtype>*>& top,
const vector<bool>& propagate_down,
const vector<Blob<Dtype>*>& bottom) {
if (!propagate_down[0]) {
return;
}
// Fallback to standard Caffe for leaky ReLU.
if (ReLULayer<Dtype>::layer_param_.relu_param().negative_slope() != 0) {
return ReLULayer<Dtype>::Backward_gpu(top, propagate_down, bottom);
}
const Dtype* top_data = top[0]->gpu_data();
const Dtype* top_diff = top[0]->gpu_diff();
const Dtype* bottom_data = bottom[0]->gpu_data();
Dtype* bottom_diff = bottom[0]->mutable_gpu_diff();
Dtype alpha = 1.0;
Dtype beta = 0.0;
CUDNN_CHECK(cudnnActivationBackward(this->handle_,
CUDNN_ACTIVATION_RELU,
reinterpret_cast<void *>(&alpha),
this->top_desc_, top_data, this->top_desc_, top_diff,
this->bottom_desc_, bottom_data,
reinterpret_cast<void *>(&beta),
this->bottom_desc_, bottom_diff));
}
INSTANTIATE_LAYER_GPU_FUNCS(CuDNNReLULayer);
} // namespace caffe
#endif
|
