-
Notifications
You must be signed in to change notification settings - Fork 0
/
scalable_go_training.cpp
284 lines (243 loc) · 11.2 KB
/
scalable_go_training.cpp
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
// Copyright [2015, 2016] <duncan@wduncanfraser.com>
#include <array>
#include <vector>
#include <iostream>
#include <string>
#include <limits>
#include <stdexcept>
#include <chrono>
#include <random>
#include "gogame.h"
#include "gogamenn.h"
#include "gogameab.h"
#define DEPTH 1
#define MUTATER 0.01
#define NETWORKCOUNT 30
#define NETWORKKEEP 10
#define BOARD_SIZE 3
#define TRAINING_SET 1
#define STARTCYCLE 1
#define ENDCYCLE 200
#define UNIFORM 0
#define SCALED 0
class TrainingArgumentError : public std::runtime_error {
public:
TrainingArgumentError() : std::runtime_error("TrainingArgumentError") { }
};
class TrainingImportError : public std::runtime_error {
public:
TrainingImportError() : std::runtime_error("TrainingImportError") { }
};
std::vector<int> score_networks(std::vector<GoGameNN> networks, const uint8_t board_size) {
// Vector to hold win counts for networks
std::vector<int> scores(networks.size(), 0);
// Each network plays every other network as each team, storing total score for each neural network.
#pragma omp parallel for firstprivate(networks) schedule(dynamic, 1)
for (unsigned int i = 0; i < networks.size(); i++) {
for (unsigned int j = 0; j < networks.size(); j++) {
// GoGame instance used for training matches
GoGame training_game(board_size);
GoMove best_move(training_game.get_board());
// Bool to determine if game should continue
bool continue_match = true;
// Value of best move
double best_move_value, temp_best_move_value = 0;
while (continue_match) {
// Generate and take black move
training_game.generate_moves(0);
best_move_value = -std::numeric_limits<double>::infinity();
// For each possible move, calculate Alpha Beta
for (const GoMove &element : training_game.get_move_list()) {
GoGame temp_game(training_game);
temp_game.make_move(element, 0);
temp_best_move_value = scalable_go_ab_prune(networks[i], temp_game, DEPTH,
-std::numeric_limits<double>::infinity(),
std::numeric_limits<double>::infinity(), 1, false, 0);
if (temp_best_move_value > best_move_value) {
best_move_value = temp_best_move_value;
best_move = element;
}
}
// Make Black Move
training_game.make_move(best_move, 0);
// Generate and take white move
training_game.generate_moves(1);
best_move_value = -std::numeric_limits<double>::infinity();
// For each possible move, calculate Alpha Beta
for (const GoMove &element : training_game.get_move_list()) {
GoGame temp_game(training_game);
temp_game.make_move(element, 1);
temp_best_move_value = scalable_go_ab_prune(networks[j], temp_game, DEPTH,
-std::numeric_limits<double>::infinity(),
std::numeric_limits<double>::infinity(), 0, false, 1);
if (temp_best_move_value > best_move_value) {
best_move_value = temp_best_move_value;
best_move = element;
}
}
// Make White move
training_game.make_move(best_move, 1);
// Game end detection
std::vector<GoMove> history(training_game.get_move_history());
// Check if the last 2 moves were passes. If so, end
if (history[history.size() - 1].check_pass() && history[history.size() - 2].check_pass()) {
std::array<uint8_t, 2> game_score = training_game.calculate_scores();
if (game_score[0] > game_score[1]) {
// Black Wins
scores[i] += 1;
scores[j] -= 1;
} else if (game_score[1] > game_score[0]) {
// White wins
scores[j] += 1;
scores[i] -= 1;
}
// Else, draw... assign no scores.
continue_match = false;
}
}
}
}
return scores;
}
int main(int argc, char* argv[]) {
std::chrono::time_point<std::chrono::system_clock> start, end;
start = std::chrono::system_clock::now();
std::time_t start_time = std::chrono::system_clock::to_time_t(start);
std::cout << "Started computation at " << std::ctime(&start_time) << std::endl;
uint8_t board_size = 0;
unsigned int training_set = 0;
unsigned int start_cycle = 0;
unsigned int end_cycle = 0;
bool uniform = 0;
bool scaled = 0;
// Validate command line parameters
if (argc == 1) {
// No parameters, use the Macros
board_size = BOARD_SIZE;
training_set = TRAINING_SET;
start_cycle = STARTCYCLE;
end_cycle = ENDCYCLE;
uniform = UNIFORM;
scaled = SCALED;
} else if (argc == 7) {
// TODO(wdfraser): Add some better error checking
board_size = uint8_t(atoi(argv[1]));
training_set = atoi(argv[2]);
start_cycle = atoi(argv[3]);
end_cycle = atoi(argv[4]);
uniform = atoi(argv[5]) != 0;
scaled = atoi(argv[6]) != 0;
} else {
throw TrainingArgumentError();
}
for (unsigned int n = start_cycle; n <= end_cycle; n++) {
std::vector<GoGameNN> training_networks(NETWORKCOUNT, GoGameNN(board_size, uniform));
std::vector<int> training_scores(NETWORKCOUNT);
// Scaling networks used for seeding networks if scaled = true
std::vector<GoGameNN> scaling_networks;
// Set Random generator for use when selecting networks for reseeding
std::random_device rd;
std::mt19937 gen(rd());
std::uniform_int_distribution<> dis(0, NETWORKKEEP - 1);
unsigned int export_count = 0;
std::string output_directory =
"size" + std::to_string(board_size) + "set" + std::to_string(training_set) + "/";
std::ifstream best_networks_in(output_directory + "lastbestnetworks.txt");
std::ifstream import_networks(output_directory + "import_networks.txt");
if (scaled) {
// If scaled network, confirm import_networks is open
if (!import_networks.is_open()) {
throw TrainingImportError();
} else {
scaling_networks.assign(NETWORKKEEP, GoGameNN(board_size - SEGMENT_DIVISION, uniform));
// If it is, import example_networks;
for (unsigned int i = 0; i < NETWORKKEEP; i++) {
scaling_networks[i].import_weights_stream(import_networks);
}
}
}
std::cout << "Generation with " << NETWORKCOUNT << " Neural Networks.\n"
<< "Total Games: " << NETWORKCOUNT * NETWORKCOUNT << std::endl;
if (best_networks_in.is_open()) {
std::cout << "Starting generation " << n << ". Last best network file succesfully opened. \n";
// Read kept networks from file
for (unsigned int i = 0; i < NETWORKKEEP; i++) {
training_networks[i].import_weights_stream(best_networks_in);
}
for (unsigned int i = 0; i < NETWORKKEEP; i++) {
training_networks[i + NETWORKKEEP] = training_networks[i];
training_networks[i + NETWORKKEEP].mutate(MUTATER);
}
for (unsigned int i = 0; i < NETWORKKEEP; i++) {
if (scaled) {
// If scaled network, seed network subsections from a random imported network
training_networks[i + (NETWORKKEEP * 2)].scale_network(scaling_networks[dis(gen)]);
} else {
training_networks[i + (NETWORKKEEP * 2)].initialize_random();
}
}
} else if (n == 1) {
std::cout << "Starting first generation. Last best network file failed to open."
<< " Initializing random weights. \n";
for (GoGameNN &element : training_networks) {
if (scaled) {
// If scaled network, seed network subsections from a random imported network
element.scale_network(scaling_networks[dis(gen)]);
} else {
element.initialize_random();
}
}
} else {
std::cout << "Starting generation " << n << ". Last best network file failed to open. Ending training. \n";
break;
}
training_scores = score_networks(training_networks, board_size);
for (unsigned int i = 0; i < training_scores.size(); i++) {
std::cout << "Neural Network: " << i << ". Score: " << training_scores[i] << ".\n";
}
std::ofstream output_file(output_directory + "generation" + std::to_string(n) + ".txt");
std::ofstream best_networks_file(output_directory + "lastbestnetworks.txt",
std::ofstream::out | std::ofstream::trunc);
if (output_file.is_open()) {
for (unsigned int i = 0; i < training_scores.size(); i++) {
output_file << "Neural Network: " << i << ". Score: " << training_scores[i] << ".\n";
}
output_file << std::endl;
for (unsigned int i = 0; i < training_scores.size(); i++) {
output_file << "Neural Network: " << i << ".\n";
training_networks[i].export_weights_stream(output_file);
output_file << std::endl;
}
output_file.close();
} else {
std::cout << "Error opening output file. \n";
break;
}
if (best_networks_file.is_open()) {
// Check down to lowest possible score, -NETWORKCOUNT*2
for (int i = NETWORKCOUNT - 1; i > -NETWORKCOUNT * 2; i--) {
if (export_count >= NETWORKKEEP) {
break;
}
for (unsigned int j = 0; j < NETWORKCOUNT; j++) {
if (export_count >= NETWORKKEEP) {
break;
}
if (training_scores[j] == i) {
training_networks[j].export_weights_stream(best_networks_file);
export_count += 1;
}
}
}
best_networks_file.close();
} else {
std::cout << "Error opening best networks file. \n";
break;
}
}
end = std::chrono::system_clock::now();
std::chrono::duration<double> elapsed_seconds = end - start;
std::time_t end_time = std::chrono::system_clock::to_time_t(end);
std::cout << "Finished computation at " << std::ctime(&end_time) << "elapsed time: " <<
elapsed_seconds.count() << "s" << std::endl;
}