uct/uct.c

   1 #include <assert.h>
   2 #include <pthread.h>
   3 #include <signal.h>
   4 #include <stdio.h>
   5 #include <stdlib.h>
   6 #include <string.h>
   7
   8 #define DEBUG
   9
  10 #include "debug.h"
  11 #include "board.h"
  12 #include "move.h"
  13 #include "playout.h"
  14 #include "playout/moggy.h"
  15 #include "playout/light.h"
  16 #include "random.h"
  17 #include "uct/internal.h"
  18 #include "uct/tree.h"
  19 #include "uct/uct.h"
  20
  21 struct uct_policy *policy_ucb1_init(struct uct *u, char *arg);
  22 struct uct_policy *policy_ucb1tuned_init(struct uct *u, char *arg);
  23 struct uct_policy *policy_ucb1amaf_init(struct uct *u, char *arg);
  24
  25
  26 #define MC_GAMES        80000
  27 #define MC_GAMELEN      400
  28
  29
  30 static void
  31 progress_status(struct uct *u, struct tree *t, enum stone color, int playouts)
  32 {
  33         if (!UDEBUGL(0))
  34                 return;
  35
  36         /* Best move */
  37         struct tree_node *best = u->policy->choose(u->policy, t->root, t->board, color);
  38         if (!best) {
  39                 fprintf(stderr, "... No moves left\n");
  40                 return;
  41         }
  42         fprintf(stderr, "[%d] ", playouts);
  43         fprintf(stderr, "best %f ", best->u.value);
  44
  45         /* Max depth */
  46         fprintf(stderr, "deepest % 2d ", t->max_depth - t->root->depth);
  47
  48         /* Best sequence */
  49         fprintf(stderr, "| seq ");
  50         for (int depth = 0; depth < 6; depth++) {
  51                 if (best && best->u.playouts >= 25) {
  52                         fprintf(stderr, "%3s ", coord2sstr(best->coord, t->board));
  53                         best = u->policy->choose(u->policy, best, t->board, color);
  54                 } else {
  55                         fprintf(stderr, "    ");
  56                 }
  57         }
  58
  59         /* Best candidates */
  60         fprintf(stderr, "| can ");
  61         int cans = 4;
  62         struct tree_node *can[cans];
  63         memset(can, 0, sizeof(can));
  64         best = t->root->children;
  65         while (best) {
  66                 int c = 0;
  67                 while ((!can[c] || best->u.playouts > can[c]->u.playouts) && ++c < cans);
  68                 for (int d = 0; d < c; d++) can[d] = can[d + 1];
  69                 if (c > 0) can[c - 1] = best;
  70                 best = best->sibling;
  71         }
  72         while (--cans >= 0) {
  73                 if (can[cans]) {
  74                         fprintf(stderr, "%3s(%.3f) ", coord2sstr(can[cans]->coord, t->board), can[cans]->u.value);
  75                 } else {
  76                         fprintf(stderr, "           ");
  77                 }
  78         }
  79
  80         fprintf(stderr, "\n");
  81 }
  82
  83
  84 static int
  85 uct_leaf_node(struct uct *u, struct board *b, enum stone player_color,
  86               struct playout_amafmap *amaf,
  87               struct tree *t, struct tree_node *n, enum stone node_color,
  88               char *spaces)
  89 {
  90         enum stone next_color = stone_other(node_color);
  91         if (n->u.playouts >= u->expand_p)
  92                 tree_expand_node(t, n, b, next_color, u->radar_d, u->policy,
  93                                  (next_color == player_color ? 1 : -1));
  94         if (UDEBUGL(7))
  95                 fprintf(stderr, "%s*-- UCT playout #%d start [%s] %f\n",
  96                         spaces, n->u.playouts, coord2sstr(n->coord, t->board), n->u.value);
  97
  98         int result = play_random_game(b, next_color, u->gamelen, u->playout_amaf ? amaf : NULL, u->playout);
  99         if (player_color != next_color && result >= 0)
 100                 result = !result;
 101         if (UDEBUGL(7))
 102                 fprintf(stderr, "%s -- [%d..%d] %s random playout result %d\n",
 103                         spaces, player_color, next_color, coord2sstr(n->coord, t->board), result);
 104
 105         return result;
 106 }
 107
 108 static int
 109 uct_playout(struct uct *u, struct board *b, enum stone player_color, struct tree *t)
 110 {
 111         struct board b2;
 112         board_copy(&b2, b);
 113
 114         struct playout_amafmap *amaf = NULL;
 115         if (u->policy->wants_amaf) {
 116                 amaf = calloc(1, sizeof(*amaf));
 117                 amaf->map = calloc(board_size2(&b2) + 1, sizeof(*amaf->map));
 118                 amaf->map++; // -1 is pass
 119         }
 120
 121         /* Walk the tree until we find a leaf, then expand it and do
 122          * a random playout. */
 123         struct tree_node *n = t->root;
 124         enum stone node_color = stone_other(player_color);
 125         int result;
 126         int pass_limit = (board_size(&b2) - 2) * (board_size(&b2) - 2) / 2;
 127         int passes = is_pass(b->last_move.coord);
 128         /* debug */
 129         int depth = 0;
 130         static char spaces[] = "\0                                                      ";
 131         /* /debug */
 132         if (UDEBUGL(8))
 133                 fprintf(stderr, "--- UCT walk with color %d\n", player_color);
 134
 135         while (!tree_leaf_node(n) && passes < 2) {
 136                 spaces[depth++] = ' '; spaces[depth] = 0;
 137
 138                 /* Parity is chosen already according to the child color, since
 139                  * it is applied to children. */
 140                 node_color = stone_other(node_color);
 141                 n = u->policy->descend(u->policy, t, n, (node_color == player_color ? 1 : -1), pass_limit);
 142
 143                 assert(n == t->root || n->parent);
 144                 if (UDEBUGL(7))
 145                         fprintf(stderr, "%s+-- UCT sent us to [%s:%d] %f\n",
 146                                 spaces, coord2sstr(n->coord, t->board), n->coord, n->u.value);
 147
 148                 if (amaf && n->coord >= -1 && !is_pass(n->coord)) {
 149                         if (amaf->map[n->coord] == S_NONE) {
 150                                 amaf->map[n->coord] = node_color;
 151                         } else {
 152                                 amaf_op(amaf->map[n->coord], +);
 153                         }
 154                 }
 155
 156                 struct move m = { n->coord, node_color };
 157                 int res = board_play(&b2, &m);
 158
 159                 if (res < 0 || (!is_pass(m.coord) && !group_at(&b2, m.coord)) /* suicide */
 160                     || b2.superko_violation) {
 161                         if (UDEBUGL(3)) {
 162                                 for (struct tree_node *ni = n; ni; ni = ni->parent)
 163                                         fprintf(stderr, "%s ", coord2sstr(ni->coord, t->board));
 164                                 fprintf(stderr, "deleting invalid %s node %d,%d res %d group %d spk %d\n",
 165                                         stone2str(node_color), coord_x(n->coord,b), coord_y(n->coord,b),
 166                                         res, group_at(&b2, m.coord), b2.superko_violation);
 167                         }
 168                         tree_delete_node(t, n);
 169                         result = -1;
 170                         goto end;
 171                 }
 172
 173                 if (is_pass(n->coord))
 174                         passes++;
 175                 else
 176                         passes = 0;
 177         }
 178
 179         if (tree_leaf_node(n)) {
 180                 result = uct_leaf_node(u, &b2, player_color, amaf, t, n, node_color, spaces);
 181
 182         } else { assert(passes >= 2);
 183
 184                 float score = board_official_score(&b2);
 185                 result = (player_color == S_BLACK) ? score < 0 : score > 0;
 186
 187                 if (UDEBUGL(5))
 188                         fprintf(stderr, "[%d..%d] %s p-p scoring playout result %d (W %f)\n",
 189                                 player_color, node_color, coord2sstr(n->coord, t->board), result, score);
 190                 if (UDEBUGL(6))
 191                         board_print(&b2, stderr);
 192         }
 193
 194         assert(n == t->root || n->parent);
 195         if (result >= 0)
 196                 u->policy->update(u->policy, t, n, node_color, player_color, amaf, result);
 197
 198 end:
 199         if (amaf) {
 200                 free(amaf->map - 1);
 201                 free(amaf);
 202         }
 203         board_done_noalloc(&b2);
 204         return result;
 205 }
 206
 207 static void
 208 prepare_move(struct engine *e, struct board *b, enum stone color, coord_t promote)
 209 {
 210         struct uct *u = e->data;
 211
 212         if (!b->moves && u->t) {
 213                 /* Stale state from last game */
 214                 tree_done(u->t);
 215                 u->t = NULL;
 216         }
 217
 218         if (!u->t) {
 219                 u->t = tree_init(b, color);
 220                 if (u->force_seed)
 221                         fast_srandom(u->force_seed);
 222                 if (UDEBUGL(0))
 223                         fprintf(stderr, "Fresh board with random seed %lu\n", fast_getseed());
 224                 //board_print(b, stderr);
 225                 tree_load(u->t, b, color);
 226         }
 227
 228         /* XXX: We hope that the opponent didn't suddenly play
 229          * several moves in the row. */
 230         if (!is_resign(promote) && !tree_promote_at(u->t, b, promote)) {
 231                 if (UDEBUGL(2))
 232                         fprintf(stderr, "<cannot find node to promote>\n");
 233                 /* Reset tree */
 234                 tree_done(u->t);
 235                 u->t = tree_init(b, color);
 236         }
 237 }
 238
 239 /* Set in main thread in case the playouts should stop. */
 240 static volatile sig_atomic_t halt = 0;
 241
 242 static int
 243 uct_playouts(struct uct *u, struct board *b, enum stone color, struct tree *t)
 244 {
 245         int i, games = u->games;
 246         if (t->root->children)
 247                 games -= t->root->u.playouts / 1.5;
 248         /* else this is highly read-out but dead-end branch of opening book;
 249          * we need to start from scratch; XXX: Maybe actually base the readout
 250          * count based on number of playouts of best node? */
 251         for (i = 0; i < games; i++) {
 252                 int result = uct_playout(u, b, color, t);
 253                 if (result < 0) {
 254                         /* Tree descent has hit invalid move. */
 255                         continue;
 256                 }
 257
 258                 if (i > 0 && !(i % 10000)) {
 259                         progress_status(u, t, color, i);
 260                 }
 261
 262                 if (i > 0 && !(i % 500)) {
 263                         struct tree_node *best = u->policy->choose(u->policy, t->root, b, color);
 264                         if (best && best->u.playouts >= 1500 && best->u.value >= u->loss_threshold)
 265                                 break;
 266                 }
 267
 268                 if (halt) {
 269                         if (UDEBUGL(2))
 270                                 fprintf(stderr, "<halting early, %d games skipped>\n", games - i);
 271                         break;
 272                 }
 273         }
 274
 275         progress_status(u, t, color, i);
 276         if (UDEBUGL(3))
 277                 tree_dump(t, u->dumpthres);
 278         return i;
 279 }
 280
 281 static pthread_mutex_t finish_mutex = PTHREAD_MUTEX_INITIALIZER;
 282 static pthread_cond_t finish_cond = PTHREAD_COND_INITIALIZER;
 283 static volatile int finish_thread;
 284 static pthread_mutex_t finish_serializer = PTHREAD_MUTEX_INITIALIZER;
 285
 286 struct spawn_ctx {
 287         int tid;
 288         struct uct *u;
 289         struct board *b;
 290         enum stone color;
 291         struct tree *t;
 292         unsigned long seed;
 293         int games;
 294 };
 295
 296 static void *
 297 spawn_helper(void *ctx_)
 298 {
 299         struct spawn_ctx *ctx = ctx_;
 300         /* Setup */
 301         fast_srandom(ctx->seed);
 302         /* Run */
 303         ctx->games = uct_playouts(ctx->u, ctx->b, ctx->color, ctx->t);
 304         /* Finish */
 305         pthread_mutex_lock(&finish_serializer);
 306         pthread_mutex_lock(&finish_mutex);
 307         finish_thread = ctx->tid;
 308         pthread_cond_signal(&finish_cond);
 309         pthread_mutex_unlock(&finish_mutex);
 310         return ctx;
 311 }
 312
 313 static void
 314 uct_notify_play(struct engine *e, struct board *b, struct move *m)
 315 {
 316         prepare_move(e, b, stone_other(m->color), m->coord);
 317 }
 318
 319 static coord_t *
 320 uct_genmove(struct engine *e, struct board *b, enum stone color)
 321 {
 322         struct uct *u = e->data;
 323
 324         /* Seed the tree. */
 325         prepare_move(e, b, color, resign);
 326
 327         int played_games = 0;
 328         if (!u->threads) {
 329                 played_games = uct_playouts(u, b, color, u->t);
 330         } else {
 331                 pthread_t threads[u->threads];
 332                 int joined = 0;
 333                 halt = 0;
 334                 pthread_mutex_lock(&finish_mutex);
 335                 /* Spawn threads... */
 336                 for (int ti = 0; ti < u->threads; ti++) {
 337                         struct spawn_ctx *ctx = malloc(sizeof(*ctx));
 338                         ctx->u = u; ctx->b = b; ctx->color = color;
 339                         ctx->t = tree_copy(u->t); ctx->tid = ti;
 340                         ctx->seed = fast_random(65536) + ti;
 341                         pthread_create(&threads[ti], NULL, spawn_helper, ctx);
 342                         if (UDEBUGL(2))
 343                                 fprintf(stderr, "Spawned thread %d\n", ti);
 344                 }
 345                 /* ...and collect them back: */
 346                 while (joined < u->threads) {
 347                         /* Wait for some thread to finish... */
 348                         pthread_cond_wait(&finish_cond, &finish_mutex);
 349                         /* ...and gather its remnants. */
 350                         struct spawn_ctx *ctx;
 351                         pthread_join(threads[finish_thread], (void **) &ctx);
 352                         played_games += ctx->games;
 353                         joined++;
 354                         tree_merge(u->t, ctx->t);
 355                         tree_done(ctx->t);
 356                         free(ctx);
 357                         if (UDEBUGL(2))
 358                                 fprintf(stderr, "Joined thread %d\n", finish_thread);
 359                         /* Do not get stalled by slow threads. */
 360                         if (joined >= u->threads / 2)
 361                                 halt = 1;
 362                         pthread_mutex_unlock(&finish_serializer);
 363                 }
 364                 pthread_mutex_unlock(&finish_mutex);
 365         }
 366
 367         if (UDEBUGL(2))
 368                 tree_dump(u->t, u->dumpthres);
 369
 370         struct tree_node *best = u->policy->choose(u->policy, u->t->root, b, color);
 371         if (!best) {
 372                 tree_done(u->t); u->t = NULL;
 373                 return coord_copy(pass);
 374         }
 375         if (UDEBUGL(0))
 376                 fprintf(stderr, "*** WINNER is %s (%d,%d) with score %1.4f (%d/%d:%d games)\n", coord2sstr(best->coord, b), coord_x(best->coord, b), coord_y(best->coord, b), best->u.value, best->u.playouts, u->t->root->u.playouts, played_games);
 377         if (best->u.value < u->resign_ratio && !is_pass(best->coord)) {
 378                 tree_done(u->t); u->t = NULL;
 379                 return coord_copy(resign);
 380         }
 381         tree_promote_node(u->t, best);
 382         return coord_copy(best->coord);
 383 }
 384
 385 bool
 386 uct_genbook(struct engine *e, struct board *b, enum stone color)
 387 {
 388         struct uct *u = e->data;
 389         u->t = tree_init(b, color);
 390         tree_load(u->t, b, color);
 391
 392         int i;
 393         for (i = 0; i < u->games; i++) {
 394                 int result = uct_playout(u, b, color, u->t);
 395                 if (result < 0) {
 396                         /* Tree descent has hit invalid move. */
 397                         continue;
 398                 }
 399
 400                 if (i > 0 && !(i % 10000)) {
 401                         progress_status(u, u->t, color, i);
 402                 }
 403         }
 404         progress_status(u, u->t, color, i);
 405
 406         tree_save(u->t, b, u->games / 100);
 407
 408         tree_done(u->t);
 409
 410         return true;
 411 }
 412
 413 void
 414 uct_dumpbook(struct engine *e, struct board *b, enum stone color)
 415 {
 416         struct uct *u = e->data;
 417         u->t = tree_init(b, color);
 418         tree_load(u->t, b, color);
 419         tree_dump(u->t, 0);
 420         tree_done(u->t);
 421 }
 422
 423
 424 struct uct *
 425 uct_state_init(char *arg)
 426 {
 427         struct uct *u = calloc(1, sizeof(struct uct));
 428
 429         u->debug_level = 1;
 430         u->games = MC_GAMES;
 431         u->gamelen = MC_GAMELEN;
 432         u->expand_p = 2;
 433         u->dumpthres = 1000;
 434         u->playout_amaf = false;
 435
 436         if (arg) {
 437                 char *optspec, *next = arg;
 438                 while (*next) {
 439                         optspec = next;
 440                         next += strcspn(next, ",");
 441                         if (*next) { *next++ = 0; } else { *next = 0; }
 442
 443                         char *optname = optspec;
 444                         char *optval = strchr(optspec, '=');
 445                         if (optval) *optval++ = 0;
 446
 447                         if (!strcasecmp(optname, "debug")) {
 448                                 if (optval)
 449                                         u->debug_level = atoi(optval);
 450                                 else
 451                                         u->debug_level++;
 452                         } else if (!strcasecmp(optname, "games") && optval) {
 453                                 u->games = atoi(optval);
 454                         } else if (!strcasecmp(optname, "gamelen") && optval) {
 455                                 u->gamelen = atoi(optval);
 456                         } else if (!strcasecmp(optname, "expand_p") && optval) {
 457                                 u->expand_p = atoi(optval);
 458                         } else if (!strcasecmp(optname, "radar_d") && optval) {
 459                                 /* For 19x19, it is good idea to set this to 3. */
 460                                 u->radar_d = atoi(optval);
 461                         } else if (!strcasecmp(optname, "dumpthres") && optval) {
 462                                 u->dumpthres = atoi(optval);
 463                         } else if (!strcasecmp(optname, "playout_amaf")) {
 464                                 /* Whether to include random playout moves in
 465                                  * AMAF as well. (Otherwise, only tree moves
 466                                  * are included in AMAF. Of course makes sense
 467                                  * only in connection with an AMAF policy.) */
 468                                 /* with-without: 55.5% (+-4.1) */
 469                                 if (optval && *optval == '0')
 470                                         u->playout_amaf = false;
 471                                 else
 472                                         u->playout_amaf = true;
 473                         } else if (!strcasecmp(optname, "policy") && optval) {
 474                                 char *policyarg = strchr(optval, ':');
 475                                 if (policyarg)
 476                                         *policyarg++ = 0;
 477                                 if (!strcasecmp(optval, "ucb1")) {
 478                                         u->policy = policy_ucb1_init(u, policyarg);
 479                                 } else if (!strcasecmp(optval, "ucb1tuned")) {
 480                                         u->policy = policy_ucb1tuned_init(u, policyarg);
 481                                 } else if (!strcasecmp(optval, "ucb1amaf")) {
 482                                         u->policy = policy_ucb1amaf_init(u, policyarg);
 483                                 } else {
 484                                         fprintf(stderr, "UCT: Invalid tree policy %s\n", optval);
 485                                 }
 486                         } else if (!strcasecmp(optname, "playout") && optval) {
 487                                 char *playoutarg = strchr(optval, ':');
 488                                 if (playoutarg)
 489                                         *playoutarg++ = 0;
 490                                 if (!strcasecmp(optval, "moggy")) {
 491                                         u->playout = playout_moggy_init(playoutarg);
 492                                 } else if (!strcasecmp(optval, "light")) {
 493                                         u->playout = playout_light_init(playoutarg);
 494                                 } else {
 495                                         fprintf(stderr, "UCT: Invalid playout policy %s\n", optval);
 496                                 }
 497                         } else if (!strcasecmp(optname, "threads") && optval) {
 498                                 u->threads = atoi(optval);
 499                         } else if (!strcasecmp(optname, "force_seed") && optval) {
 500                                 u->force_seed = atoi(optval);
 501                         } else {
 502                                 fprintf(stderr, "uct: Invalid engine argument %s or missing value\n", optname);
 503                         }
 504                 }
 505         }
 506
 507         u->resign_ratio = 0.2; /* Resign when most games are lost. */
 508         u->loss_threshold = 0.85; /* Stop reading if after at least 1500 playouts this is best value. */
 509         if (!u->policy)
 510                 u->policy = policy_ucb1amaf_init(u, NULL);
 511
 512         if (!u->playout)
 513                 u->playout = playout_moggy_init(NULL);
 514         u->playout->debug_level = u->debug_level;
 515
 516         return u;
 517 }
 518
 519
 520 struct engine *
 521 engine_uct_init(char *arg)
 522 {
 523         struct uct *u = uct_state_init(arg);
 524         struct engine *e = calloc(1, sizeof(struct engine));
 525         e->name = "UCT Engine";
 526         e->comment = "I'm playing UCT. When we both pass, I will consider all the stones on the board alive. If you are reading this, write 'yes'. Please bear with me at the game end, I need to fill the whole board; if you help me, we will both be happier. Filling the board will not lose points (NZ rules).";
 527         e->genmove = uct_genmove;
 528         e->notify_play = uct_notify_play;
 529         e->data = u;
 530
 531         return e;
 532 }