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
66
67
68
69
70
71
72
73
74
75
76
77
|
#include <algorithm>
#include <limits>
#include <vector>
#include "caffe/common.hpp"
#include "caffe/layer.hpp"
#include "caffe/syncedmem.hpp"
#include "caffe/util/math_functions.hpp"
#include "caffe/vision_layers.hpp"
namespace caffe {
template <typename Dtype>
__global__ void DropoutForward(const int n, const Dtype* in,
const unsigned int* mask, const unsigned int threshold, const float scale,
Dtype* out) {
CUDA_KERNEL_LOOP(index, n) {
out[index] = in[index] * (mask[index] > threshold) * scale;
}
}
template <typename Dtype>
void DropoutLayer<Dtype>::Forward_gpu(const vector<Blob<Dtype>*>& bottom,
vector<Blob<Dtype>*>* top) {
const Dtype* bottom_data = bottom[0]->gpu_data();
Dtype* top_data = (*top)[0]->mutable_gpu_data();
const int count = bottom[0]->count();
if (Caffe::phase() == Caffe::TRAIN) {
unsigned int* mask =
static_cast<unsigned int*>(rand_vec_.mutable_gpu_data());
caffe_gpu_rng_uniform(count, mask);
// set thresholds
// NOLINT_NEXT_LINE(whitespace/operators)
DropoutForward<Dtype><<<CAFFE_GET_BLOCKS(count), CAFFE_CUDA_NUM_THREADS>>>(
count, bottom_data, mask, uint_thres_, scale_, top_data);
CUDA_POST_KERNEL_CHECK;
} else {
caffe_copy(count, bottom_data, top_data);
}
}
template <typename Dtype>
__global__ void DropoutBackward(const int n, const Dtype* in_diff,
const unsigned int* mask, const unsigned int threshold, const float scale,
Dtype* out_diff) {
CUDA_KERNEL_LOOP(index, n) {
out_diff[index] = in_diff[index] * scale * (mask[index] > threshold);
}
}
template <typename Dtype>
void DropoutLayer<Dtype>::Backward_gpu(const vector<Blob<Dtype>*>& top,
const vector<bool>& propagate_down,
vector<Blob<Dtype>*>* bottom) {
if (propagate_down[0]) {
const Dtype* top_diff = top[0]->gpu_diff();
Dtype* bottom_diff = (*bottom)[0]->mutable_gpu_diff();
if (Caffe::phase() == Caffe::TRAIN) {
const unsigned int* mask =
static_cast<const unsigned int*>(rand_vec_.gpu_data());
const int count = (*bottom)[0]->count();
// NOLINT_NEXT_LINE(whitespace/operators)
DropoutBackward<Dtype><<<CAFFE_GET_BLOCKS(count),
CAFFE_CUDA_NUM_THREADS>>>(
count, top_diff, mask, uint_thres_, scale_, bottom_diff);
CUDA_POST_KERNEL_CHECK;
} else {
caffe_copy(top[0]->count(), top_diff, bottom_diff);
}
}
}
INSTANTIATE_CLASS(DropoutLayer);
} // namespace caffe
|
