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
|
/* Copyright 2017 The TensorFlow Authors. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
==============================================================================*/
// XLA implementation of BatchNorm operations.
#include "tensorflow/compiler/tf2xla/type_util.h"
#include "tensorflow/compiler/tf2xla/xla_helpers.h"
#include "tensorflow/compiler/tf2xla/xla_op_kernel.h"
#include "tensorflow/compiler/tf2xla/xla_op_registry.h"
#include "tensorflow/compiler/xla/client/xla_builder.h"
#include "tensorflow/core/util/tensor_format.h"
namespace tensorflow {
namespace {
class FusedBatchNormOp : public XlaOpKernel {
public:
explicit FusedBatchNormOp(OpKernelConstruction* ctx) : XlaOpKernel(ctx) {
OP_REQUIRES_OK(ctx, ctx->GetAttr("epsilon", &epsilon_));
OP_REQUIRES_OK(ctx, ctx->GetAttr("is_training", &is_training_));
string data_format_str;
OP_REQUIRES_OK(ctx, ctx->GetAttr("data_format", &data_format_str));
OP_REQUIRES(
ctx, FormatFromString(data_format_str, &data_format_),
errors::InvalidArgument("Invalid data format: ", data_format_str));
}
void Compile(XlaOpKernelContext* ctx) override {
xla::PrimitiveType input_type;
OP_REQUIRES_OK(ctx,
DataTypeToPrimitiveType(ctx->input_type(0), &input_type));
xla::PrimitiveType scale_type;
OP_REQUIRES_OK(ctx,
DataTypeToPrimitiveType(ctx->input_type(1), &scale_type));
xla::XlaOp input = ctx->Input(0);
TensorShape input_shape = ctx->InputShape(0);
int feature_index =
GetTensorFeatureDimIndex(input_shape.dims(), data_format_);
// TODO(b/69928690): support mixed precision in the XLA batch normalization
// operators. As a workaround, cast everything to the statistics type (which
// may be more precise than the input type).
input = xla::ConvertElementType(input, scale_type);
if (is_training_) {
xla::XlaOp output = xla::BatchNormTraining(
input, ctx->Input(1), ctx->Input(2), epsilon_, feature_index);
// In training mode, outputs the normalized value as well as the
// calculated mean and variance.
ctx->SetOutput(0, xla::ConvertElementType(xla::GetTupleElement(output, 0),
input_type));
ctx->SetOutput(1, xla::GetTupleElement(output, 1));
ctx->SetOutput(2, xla::GetTupleElement(output, 2));
// Output 3 and 4 for "FusedBatchNorm" are currently marked as "reserved
// space 1 & 2". They are used to pass the per-batch mean and
// variance to the gradient. Here we maintain the same behavior by setting
// them to the mean and variance calculated by BatchNormTraining.
ctx->SetOutput(3, xla::GetTupleElement(output, 1));
ctx->SetOutput(4, xla::GetTupleElement(output, 2));
} else {
xla::XlaOp output = xla::BatchNormInference(
input, ctx->Input(1), ctx->Input(2), ctx->Input(3), ctx->Input(4),
epsilon_, feature_index);
ctx->SetOutput(0, xla::ConvertElementType(output, input_type));
// Directly send input to output as mean and variance in inference mode.
ctx->SetOutput(1, ctx->Input(3));
ctx->SetOutput(2, ctx->Input(4));
ctx->SetOutput(3, ctx->Input(3));
ctx->SetOutput(4, ctx->Input(4));
}
}
private:
float epsilon_;
TensorFormat data_format_;
bool is_training_;
};
REGISTER_XLA_OP(Name("FusedBatchNorm"), FusedBatchNormOp);
REGISTER_XLA_OP(Name("FusedBatchNormV2"), FusedBatchNormOp);
class FusedBatchNormGradOp : public XlaOpKernel {
public:
explicit FusedBatchNormGradOp(OpKernelConstruction* ctx) : XlaOpKernel(ctx) {
OP_REQUIRES_OK(ctx, ctx->GetAttr("epsilon", &epsilon_));
OP_REQUIRES_OK(ctx, ctx->GetAttr("is_training", &is_training_));
string data_format_str;
OP_REQUIRES_OK(ctx, ctx->GetAttr("data_format", &data_format_str));
OP_REQUIRES(
ctx, FormatFromString(data_format_str, &data_format_),
errors::InvalidArgument("Invalid data format: ", data_format_str));
}
void Compile(XlaOpKernelContext* ctx) override {
xla::XlaBuilder* const b = ctx->builder();
DataType input_dtype = ctx->input_type(0);
DataType scale_dtype = ctx->input_type(2);
// TODO(b/69928690): support mixed precision in the XLA batch normalization
// operators. For now, cast everything to the statistics type (which
// may be more precise than the input type).
auto grad_backprop =
XlaHelpers::ConvertElementType(b, ctx->Input(0), scale_dtype);
auto activations =
XlaHelpers::ConvertElementType(b, ctx->Input(1), scale_dtype);
auto scale = ctx->Input(2);
auto mean = ctx->Input(3);
auto var = ctx->Input(4);
const int input_dims = ctx->InputShape(0).dims();
const int feature_index =
GetTensorFeatureDimIndex(input_dims, data_format_);
xla::XlaOp x_backprop;
xla::XlaOp scale_backprop;
xla::XlaOp offset_backprop;
if (is_training_) {
xla::XlaOp output =
xla::BatchNormGrad(activations, scale, mean, var, grad_backprop,
epsilon_, feature_index);
x_backprop = xla::GetTupleElement(output, 0);
scale_backprop = xla::GetTupleElement(output, 1);
offset_backprop = xla::GetTupleElement(output, 2);
} else {
// Reduce over all dimensions except the feature dim.
std::vector<int64> reduction_dims(input_dims - 1);
std::iota(reduction_dims.begin(), reduction_dims.begin() + feature_index,
0);
std::iota(reduction_dims.begin() + feature_index, reduction_dims.end(),
feature_index + 1);
// offset_backprop = sum(y_backprop)
// scale_backprop = y_backprop * ((x - pop_mean) * rsqrt(pop_var +
// epsilon))
// x_backprop = y_backprop * (scale * rsqrt(pop_var + epsilon))
const DataType accumulation_type =
XlaHelpers::SumAccumulationType(scale_dtype);
auto converted =
XlaHelpers::ConvertElementType(b, grad_backprop, accumulation_type);
auto reduce =
xla::Reduce(converted, XlaHelpers::Zero(b, accumulation_type),
*ctx->GetOrCreateAdd(accumulation_type), reduction_dims);
offset_backprop = XlaHelpers::ConvertElementType(b, reduce, scale_dtype);
// scratch1 = rsqrt(pop_var + epsilon)
auto neg_half = XlaHelpers::FloatLiteral(b, scale_dtype, -0.5);
auto scratch1 = xla::Pow(
xla::Add(var, xla::ConstantR0<float>(b, epsilon_)), neg_half);
// scratch2 = sum(y_backprop * (x - mean))
auto mul =
xla::Mul(grad_backprop, xla::Sub(activations, mean, {feature_index}));
converted = XlaHelpers::ConvertElementType(b, mul, accumulation_type);
reduce =
xla::Reduce(converted, XlaHelpers::Zero(b, accumulation_type),
*ctx->GetOrCreateAdd(accumulation_type), reduction_dims);
auto scratch2 = XlaHelpers::ConvertElementType(b, reduce, scale_dtype);
x_backprop =
xla::Mul(grad_backprop, xla::Mul(scratch1, scale), {feature_index});
scale_backprop = xla::Mul(scratch1, scratch2);
}
ctx->SetOutput(0,
XlaHelpers::ConvertElementType(b, x_backprop, input_dtype));
ctx->SetOutput(1, scale_backprop);
ctx->SetOutput(2, offset_backprop);
ctx->SetConstantOutput(3, Tensor());
ctx->SetConstantOutput(4, Tensor());
}
private:
TensorFormat data_format_;
float epsilon_;
bool is_training_;
};
REGISTER_XLA_OP(Name("FusedBatchNormGrad"), FusedBatchNormGradOp);
REGISTER_XLA_OP(Name("FusedBatchNormGradV2"), FusedBatchNormGradOp);
} // namespace
} // namespace tensorflow
|
