forked from pytorch/pytorch
-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathelementwise_mul_gradient_op.cc
347 lines (316 loc) · 10.1 KB
/
elementwise_mul_gradient_op.cc
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
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
#include <c10/util/accumulate.h>
#include "caffe2/operators/elementwise_mul_op.h"
#include "caffe2/utils/math/broadcast.h"
#include <algorithm>
#include <functional>
#include <string>
#include <vector>
namespace caffe2 {
namespace {
template <typename TGrad, typename TIn>
void ComputeMulGradientFastpath(
const int A_size,
const int B_size,
const int C_size,
const TGrad* dC,
const TIn* A,
const TIn* B,
TGrad* dA,
TGrad* dB) {
int A_index = 0;
int B_index = 0;
for (int C_index = 0; C_index < C_size; ++C_index) {
dA[A_index] += dC[C_index] * B[B_index];
dB[B_index] += dC[C_index] * A[A_index];
A_index++;
B_index++;
if (A_index >= A_size) {
A_index = 0;
}
if (B_index >= B_size) {
B_index = 0;
}
}
}
template <typename TGrad, typename TIn>
void ComputeMulGradient(
const int ndim,
const int* A_dims,
const int* B_dims,
const int* C_dims,
const TGrad* dC,
const TIn* A,
const TIn* B,
TGrad* dA,
TGrad* dB,
CPUContext* context) {
const auto A_size = c10::multiply_integers(A_dims, A_dims + ndim);
const auto B_size = c10::multiply_integers(B_dims, B_dims + ndim);
const auto C_size = c10::multiply_integers(C_dims, C_dims + ndim);
math::Set<TGrad, CPUContext>(A_size, TGrad(0), dA, context);
math::Set<TGrad, CPUContext>(B_size, TGrad(0), dB, context);
if (
math::can_use_broadcast_fastpath(ndim, A_dims)
&& math::can_use_broadcast_fastpath(ndim, B_dims)) {
ComputeMulGradientFastpath(A_size, B_size, C_size, dC, A, B, dA, dB);
return;
}
std::vector<int> index(ndim, 0);
for (int C_index = 0; C_index < C_size; ++C_index) {
const int A_index =
math::utils::GetIndexFromDims(ndim, A_dims, index.data());
const int B_index =
math::utils::GetIndexFromDims(ndim, B_dims, index.data());
dA[A_index] += dC[C_index] * B[B_index];
dB[B_index] += dC[C_index] * A[A_index];
math::utils::IncreaseIndexInDims(ndim, C_dims, index.data());
}
}
// A : input not to broadcast whose size is common_size x broadcast_size
// B : input to broadcast whose size is common_size
void ComputeMulGradient(
const int common_size,
const int broadcast_size,
const float* dC,
const float* A,
const float* B,
float* dA,
float* dB,
CPUContext* context) {
for (int i = 0; i < common_size; ++i) {
caffe2::math::Scale(
broadcast_size,
B[i],
dC + i * broadcast_size,
dA + i * broadcast_size,
context);
caffe2::math::Dot(
broadcast_size,
dC + i * broadcast_size,
A + i * broadcast_size,
dB + i,
context);
}
}
void ComputeMulGradient(
const int size,
const float* dC,
const float* A,
const float* B,
float* dA,
float* dB) {
if (dA != nullptr) {
CAFFE_ENFORCE_NE(dA, dB, "Outputs dA and dB should point to distinct blobs");
}
if (dC == dA) {
// Ensure operation can be performed in-place.
// See below comment in `MulFunctor::Backward`.
std::swap(A, B);
std::swap(dA, dB);
}
for (int i = 0; i < size; ++i) {
dA[i] = dC[i] * B[i];
dB[i] = dC[i] * A[i];
}
}
} // namespace
template <>
template <typename TGrad, typename TIn, typename TOut>
bool MulFunctor<CPUContext>::Backward(
const std::vector<int>& A_dims,
const std::vector<int>& B_dims,
const TGrad* dC,
const TIn* A,
const TIn* B,
const TOut* /* C */,
TGrad* dA,
TGrad* dB,
CPUContext* context) const {
if (dA != nullptr) {
CAFFE_ENFORCE_NE(dA, dB, "Outputs dA and dB should point to distinct blobs");
}
if (A_dims == B_dims) {
const auto size = c10::multiply_integers(A_dims);
if (dC == dA) {
// A, B, and dC are inputs (dC is the output of the previous gradient op
// in the dag), and dA and dB are outputs. If the op is performed
// in-place, either dA or dB could alias dC. In the dC == dA case, we need
// to make sure we don't overwrite dC when we write to dA, so swap the
// inputs to avoid clobbering dC. Semantically this is equivalent with
// writing to dB first. The other case (dC == dB) is already safe because
// we are writing to dA first.
std::swap(A, B);
std::swap(dA, dB);
}
math::Mul(size, dC, B, dA, context);
math::Mul(size, dC, A, dB, context);
return true;
}
const int ndim = std::max(A_dims.size(), B_dims.size());
if (ndim == 0) {
return true;
}
std::vector<int> A_broadcast_dims(ndim);
std::vector<int> B_broadcast_dims(ndim);
std::vector<int> C_broadcast_dims(ndim);
math::utils::ComputeBroadcastBinaryOpDims(
A_dims.size(),
A_dims.data(),
B_dims.size(),
B_dims.data(),
A_broadcast_dims.data(),
B_broadcast_dims.data(),
C_broadcast_dims.data());
const int C_size = std::accumulate(
C_broadcast_dims.cbegin(),
C_broadcast_dims.cbegin() + ndim,
1,
// NOLINTNEXTLINE(modernize-use-transparent-functors)
std::multiplies<int>());
if (C_size == 0) {
const auto A_size = c10::multiply_integers(A_dims);
const auto B_size = c10::multiply_integers(B_dims);
math::Set<TGrad, CPUContext>(A_size, TGrad(0), dA, context);
math::Set<TGrad, CPUContext>(B_size, TGrad(0), dB, context);
return true;
}
// Flatten dims as much as possible
// We call A is broadcasted at dim d if A_broadcast_dims[d] <= 1
// Two consecutive dims d and d+1 can be flattened if
// A and B are broadcasted at dim d, or
// A and B are broadcasted at dim d + 1, or
// A is broadcasted at dim d and d + 1, or
// B is broadcasted at dim d and d + 1, or
// A and B are not broadcasted at dim d and d + 1
std::vector<int> A_broadcast_dims_flattened, B_broadcast_dims_flattened,
C_broadcast_dims_flattened;
A_broadcast_dims_flattened.reserve(ndim);
B_broadcast_dims_flattened.reserve(ndim);
A_broadcast_dims_flattened.push_back(A_broadcast_dims[0]);
B_broadcast_dims_flattened.push_back(B_broadcast_dims[0]);
for (int i = 1; i < ndim; ++i) {
int A_old = A_broadcast_dims_flattened.back();
int B_old = B_broadcast_dims_flattened.back();
int A_new = A_broadcast_dims[i];
int B_new = B_broadcast_dims[i];
if ((A_old == 1 && B_old == 1) || (A_new == 1 && B_new == 1) ||
(A_old == 1 && A_new == 1) || (B_old == 1 && B_new == 1) ||
(A_old > 1 && B_old > 1 && A_new > 1 && B_new > 1)) {
A_broadcast_dims_flattened.back() *= A_new;
B_broadcast_dims_flattened.back() *= B_new;
} else {
A_broadcast_dims_flattened.push_back(A_new);
B_broadcast_dims_flattened.push_back(B_new);
}
}
int ndim_flattened = A_broadcast_dims_flattened.size();
C_broadcast_dims_flattened.resize(ndim_flattened);
for (int i = 0; i < ndim_flattened; ++i) {
C_broadcast_dims_flattened[i] =
std::max(A_broadcast_dims_flattened[i], B_broadcast_dims_flattened[i]);
}
if (std::is_same<TGrad, float>::value && std::is_same<TIn, float>::value &&
ndim_flattened <= 2 &&
A_broadcast_dims_flattened[0] == B_broadcast_dims_flattened[0] &&
(ndim_flattened == 1 || A_broadcast_dims_flattened[1] <= 1 ||
B_broadcast_dims_flattened[1] <= 1)) {
if (ndim_flattened == 2) {
// fast path when we have 2 flattened dimensions and the second dimension
// is broadcasted.
bool broadcast_B = B_broadcast_dims_flattened[1] <= 1;
ComputeMulGradient(
C_broadcast_dims_flattened[0],
C_broadcast_dims_flattened[1],
reinterpret_cast<const float*>(dC),
reinterpret_cast<const float*>(broadcast_B ? A : B),
reinterpret_cast<const float*>(broadcast_B ? B : A),
reinterpret_cast<float*>(broadcast_B ? dA : dB),
reinterpret_cast<float*>(broadcast_B ? dB : dA),
context);
} else {
// fast path when we have 1 flattened dimension
assert(ndim_flattened == 1);
ComputeMulGradient(
C_broadcast_dims_flattened[0],
reinterpret_cast<const float*>(dC),
reinterpret_cast<const float*>(A),
reinterpret_cast<const float*>(B),
reinterpret_cast<float*>(dA),
reinterpret_cast<float*>(dB));
}
} else {
ComputeMulGradient<TGrad, TIn>(
ndim_flattened,
A_broadcast_dims_flattened.data(),
B_broadcast_dims_flattened.data(),
C_broadcast_dims_flattened.data(),
dC,
A,
B,
dA,
dB,
context);
}
return true;
}
// Used in fallback ops
template bool MulFunctor<CPUContext>::Backward<float, float, float>(
const std::vector<int>& A_dims,
const std::vector<int>& B_dims,
const float* dC,
const float* A,
const float* B,
const float* /* C */,
float* dA,
float* dB,
CPUContext* context) const;
template bool MulFunctor<CPUContext>::Backward<int32_t, int32_t, int32_t>(
const std::vector<int>& A_dims,
const std::vector<int>& B_dims,
const int* dC,
const int* A,
const int* B,
const int* /* C */,
int* dA,
int* dB,
CPUContext* context) const;
template bool MulFunctor<CPUContext>::Backward<double, double, double>(
const std::vector<int>& A_dims,
const std::vector<int>& B_dims,
const double* dC,
const double* A,
const double* B,
const double* /* C */,
double* dA,
double* dB,
CPUContext* context) const;
template bool MulFunctor<CPUContext>::Backward<int64_t, int64_t, int64_t>(
const std::vector<int>& A_dims,
const std::vector<int>& B_dims,
const int64_t* dC,
const int64_t* A,
const int64_t* B,
const int64_t* /* C */,
int64_t* dA,
int64_t* dB,
CPUContext* context) const;
REGISTER_CPU_OPERATOR(
MulGradient,
BinaryElementwiseGradientOp<
NumericTypes,
CPUContext,
MulFunctor<CPUContext>>);
namespace {
class GetMulGradient final : public GradientMakerBase {
using GradientMakerBase::GradientMakerBase;
std::vector<OperatorDef> GetGradientDefs() override {
return SingleGradientDef(
"MulGradient",
"",
std::vector<std::string>{GO(0), I(0), I(1)},
std::vector<std::string>{GI(0), GI(1)});
}
};
} // namespace
REGISTER_GRADIENT(Mul, GetMulGradient);
} // namespace caffe2