Repository navigation
Expand file tree
/
Copy pathtf_utils.cc
More file actions
147 lines (118 loc) · 4.73 KB
/
Copy pathtf_utils.cc
File metadata and controls
147 lines (118 loc) · 4.73 KB
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
// Copyright 2018 Google LLC
//
// 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.
#include "cc/tf_utils.h"
#include <algorithm>
#include <array>
#include <memory>
#include "cc/constants.h"
#include "cc/file/path.h"
#include "cc/file/utils.h"
#include "cc/model/model.h"
#include "tensorflow/core/example/example.pb.h"
#include "tensorflow/core/lib/core/status.h"
#include "tensorflow/core/lib/io/record_writer.h"
#include "tensorflow/core/platform/env.h"
#include "tensorflow/core/platform/file_system.h"
using tensorflow::io::RecordWriter;
using tensorflow::io::RecordWriterOptions;
namespace minigo {
namespace tf_utils {
namespace {
tensorflow::Feature MakeBytesFeature(const Tensor<uint8_t>& src) {
int size = src.shape.num_elements();
tensorflow::Feature feature;
feature.mutable_bytes_list()->add_value(
reinterpret_cast<const void*>(src.data), size);
return feature;
}
template <size_t N>
tensorflow::Feature MakeBytesFeature(const std::array<float, N>& data) {
tensorflow::Feature feature;
feature.mutable_bytes_list()->add_value(
reinterpret_cast<const void*>(data.data()), sizeof(float) * data.size());
return feature;
}
// Converts board features, and the pi & value outputs of MTCS to a tensorflow
// example proto.
tensorflow::Example MakeTfExample(const Tensor<uint8_t>& features,
const std::array<float, kNumMoves>& pi,
float Q, int N, Coord c, float outcome) {
tensorflow::Example example;
auto& dst_features = *example.mutable_features()->mutable_feature();
// The input features are expected to be uint8 bytes.
dst_features["x"] = MakeBytesFeature(features);
// pi is expected to be a float array serialized as bytes.
dst_features["pi"] = MakeBytesFeature(pi);
// outcome is a single float.
dst_features["outcome"].mutable_float_list()->add_value(outcome);
// Q is a single float.
dst_features["q"].mutable_float_list()->add_value(Q);
// Number of reads is a single int.
dst_features["n"].mutable_int64_list()->add_value(N);
// The move played is a single int.
dst_features["c"].mutable_int64_list()->add_value(c);
return example;
}
// Writes a list of tensorflow Example protos to a zlib compressed TFRecord
// file.
void WriteTfExamples(const std::string& path,
const std::vector<tensorflow::Example>& examples) {
std::unique_ptr<tensorflow::WritableFile> file;
TF_CHECK_OK(tensorflow::Env::Default()->NewWritableFile(path, &file));
RecordWriterOptions options;
options.compression_type = RecordWriterOptions::ZLIB_COMPRESSION;
options.zlib_options.compression_level = 2;
RecordWriter writer(file.get(), options);
std::string data;
for (const auto& example : examples) {
example.SerializeToString(&data);
TF_CHECK_OK(writer.WriteRecord(data));
}
TF_CHECK_OK(writer.Close());
TF_CHECK_OK(file->Close());
}
} // namespace
std::vector<tensorflow::Example> MakeExamples(
const FeatureDescriptor& feature_desc, const Game& game) {
// Write the TensorFlow examples.
std::vector<tensorflow::Example> examples;
examples.reserve(game.num_moves());
auto shape = feature_desc.GetInputShape(1);
BoardFeatureBuffer<uint8_t> features_buffer;
Tensor<uint8_t> features(shape, features_buffer.data());
for (size_t i = 0; i < game.moves().size(); ++i) {
const auto* move = game.moves()[i].get();
if (!move->is_trainable()) {
continue;
}
ModelInput input;
input.sym = symmetry::kIdentity;
game.GetPositionHistory(i, kMaxPositionHistory, &input.position_history);
feature_desc.set_bytes({&input}, &features);
examples.push_back(MakeTfExample(features, move->search_pi.value(), move->Q,
move->N, move->c, game.result()));
}
return examples;
}
void WriteGameExamples(const std::string& output_dir,
const std::string& output_name,
const FeatureDescriptor& feature_desc,
const Game& game) {
MG_CHECK(file::RecursivelyCreateDir(output_dir));
auto output_path = file::JoinPath(output_dir, output_name + ".tfrecord.zz");
auto examples = MakeExamples(feature_desc, game);
WriteTfExamples(output_path, examples);
}
} // namespace tf_utils
} // namespace minigo