Repository navigation
Expand file tree
/
Copy pathtpu_dual_net.h
More file actions
84 lines (67 loc) · 2.53 KB
/
Copy pathtpu_dual_net.h
File metadata and controls
84 lines (67 loc) · 2.53 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
// 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.
#ifndef CC_DUAL_NET_TPU_DUAL_NET_H_
#define CC_DUAL_NET_TPU_DUAL_NET_H_
#include <memory>
#include <string>
#include <utility>
#include <vector>
#include "cc/constants.h"
#include "cc/dual_net/dual_net.h"
#include "cc/thread_safe_queue.h"
#include "tensorflow/core/framework/graph.pb.h"
#include "tensorflow/core/framework/tensor.h"
#include "tensorflow/core/public/session.h"
namespace minigo {
class TpuDualNet : public DualNet {
public:
// tpu_name: The gRPC address of the VM's TPU, e.g. "grpc://10.240.2.10:8470".
// graph_path: A frozen GraphDef proto suitable for running on TPU, e.g.
// generated by freeze_graph.py --use_tpu
TpuDualNet(const std::string& tpu_name, const std::string& graph_path);
~TpuDualNet() override;
void RunMany(std::vector<const BoardFeatures*> features,
std::vector<Output*> outputs, std::string* model) override;
private:
class Worker {
public:
Worker(const tensorflow::GraphDef& graph_def, const std::string& tpu_name,
int num_replicas);
~Worker();
void RunMany(std::vector<const DualNet::BoardFeatures*> features,
std::vector<DualNet::Output*> outputs);
private:
void Reserve(size_t capacity);
std::unique_ptr<tensorflow::Session> session_;
std::vector<std::pair<std::string, tensorflow::Tensor>> inputs_;
std::vector<std::string> output_names_;
std::vector<tensorflow::Tensor> outputs_;
const int num_replicas_;
size_t batch_capacity_;
};
ThreadSafeQueue<std::unique_ptr<Worker>> workers_;
std::string graph_path_;
};
class TpuDualNetFactory : public DualNetFactory {
public:
explicit TpuDualNetFactory(std::string tpu_name);
~TpuDualNetFactory() override;
int GetBufferCount() const override;
std::unique_ptr<DualNet> NewDualNet(const std::string& model) override;
private:
std::unique_ptr<tensorflow::Session> main_session_;
std::string tpu_name_;
};
} // namespace minigo
#endif // CC_DUAL_NET_TPU_DUAL_NET_H_