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                 // fprintf(stderr, "expanding %s (%p ^-%p)\n", coord2sstr(n->coord, b), n, n->parent);
  93                 tree_expand_node(t, n, b, next_color, u->radar_d, u->policy,
  94                                  (next_color == player_color ? 1 : -1));
  95         }
  96         if (UDEBUGL(7))
  97                 fprintf(stderr, "%s*-- UCT playout #%d start [%s] %f\n",
  98                         spaces, n->u.playouts, coord2sstr(n->coord, t->board), n->u.value);
  99
 100         int result = play_random_game(b, next_color, u->gamelen, u->playout_amaf ? amaf : NULL, u->playout);
 101         if (player_color != next_color && result >= 0)
 102                 result = !result;
 103         if (UDEBUGL(7))
 104                 fprintf(stderr, "%s -- [%d..%d] %s random playout result %d\n",
 105                         spaces, player_color, next_color, coord2sstr(n->coord, t->board), result);
 106
 107         return result;
 108 }
 109
 110 static int
 111 uct_playout(struct uct *u, struct board *b, enum stone player_color, struct tree *t)
 112 {
 113         struct board b2;
 114         board_copy(&b2, b);
 115
 116         struct playout_amafmap *amaf = NULL;
 117         if (u->policy->wants_amaf) {
 118                 amaf = calloc(1, sizeof(*amaf));
 119                 amaf->map = calloc(board_size2(&b2) + 1, sizeof(*amaf->map));
 120                 amaf->map++; // -1 is pass
 121         }
 122
 123         /* Walk the tree until we find a leaf, then expand it and do
 124          * a random playout. */
 125         struct tree_node *n = t->root;
 126         enum stone node_color = stone_other(player_color);
 127         int result;
 128         int pass_limit = (board_size(&b2) - 2) * (board_size(&b2) - 2) / 2;
 129         int passes = is_pass(b->last_move.coord);
 130         /* debug */
 131         int depth = 0;
 132         static char spaces[] = "\0                                                      ";
 133         /* /debug */
 134         if (UDEBUGL(8))
 135                 fprintf(stderr, "--- UCT walk with color %d\n", player_color);
 136
 137         while (!tree_leaf_node(n) && passes < 2) {
 138                 spaces[depth++] = ' '; spaces[depth] = 0;
 139
 140                 /* Parity is chosen already according to the child color, since
 141                  * it is applied to children. */
 142                 node_color = stone_other(node_color);
 143                 n = u->policy->descend(u->policy, t, n, (node_color == player_color ? 1 : -1), pass_limit);
 144
 145                 assert(n == t->root || n->parent);
 146                 if (UDEBUGL(7))
 147                         fprintf(stderr, "%s+-- UCT sent us to [%s:%d] %f\n",
 148                                 spaces, coord2sstr(n->coord, t->board), n->coord, n->u.value);
 149
 150                 if (amaf && n->coord >= -1 && !is_pass(n->coord)) {
 151                         if (amaf->map[n->coord] == S_NONE) {
 152                                 amaf->map[n->coord] = node_color;
 153                         } else {
 154                                 amaf_op(amaf->map[n->coord], +);
 155                         }
 156                 }
 157
 158                 struct move m = { n->coord, node_color };
 159                 int res = board_play(&b2, &m);
 160
 161                 if (res < 0 || (!is_pass(m.coord) && !group_at(&b2, m.coord)) /* suicide */
 162                     || b2.superko_violation) {
 163                         if (UDEBUGL(3)) {
 164                                 for (struct tree_node *ni = n; ni; ni = ni->parent)
 165                                         fprintf(stderr, "%s ", coord2sstr(ni->coord, t->board));
 166                                 fprintf(stderr, "deleting invalid %s node %d,%d res %d group %d spk %d\n",
 167                                         stone2str(node_color), coord_x(n->coord,b), coord_y(n->coord,b),
 168                                         res, group_at(&b2, m.coord), b2.superko_violation);
 169                         }
 170                         tree_delete_node(t, n);
 171                         result = -1;
 172                         goto end;
 173                 }
 174
 175                 if (is_pass(n->coord))
 176                         passes++;
 177                 else
 178                         passes = 0;
 179         }
 180
 181         if (tree_leaf_node(n)) {
 182                 result = uct_leaf_node(u, &b2, player_color, amaf, t, n, node_color, spaces);
 183
 184         } else { assert(passes >= 2);
 185
 186                 float score = board_official_score(&b2);
 187                 result = (player_color == S_BLACK) ? score < 0 : score > 0;
 188
 189                 if (UDEBUGL(5))
 190                         fprintf(stderr, "[%d..%d] %s p-p scoring playout result %d (W %f)\n",
 191                                 player_color, node_color, coord2sstr(n->coord, t->board), result, score);
 192                 if (UDEBUGL(6))
 193                         board_print(&b2, stderr);
 194         }
 195
 196         assert(n == t->root || n->parent);
 197         if (result >= 0)
 198                 u->policy->update(u->policy, t, n, node_color, player_color, amaf, result);
 199
 200 end:
 201         if (amaf) {
 202                 free(amaf->map - 1);
 203                 free(amaf);
 204         }
 205         board_done_noalloc(&b2);
 206         return result;
 207 }
 208
 209 static void
 210 prepare_move(struct engine *e, struct board *b, enum stone color, coord_t promote)
 211 {
 212         struct uct *u = e->data;
 213
 214         if (!b->moves && u->t) {
 215                 /* Stale state from last game */
 216                 tree_done(u->t);
 217                 u->t = NULL;
 218         }
 219
 220         if (!u->t) {
 221                 u->t = tree_init(b, color);
 222                 if (u->force_seed)
 223                         fast_srandom(u->force_seed);
 224                 if (UDEBUGL(0))
 225                         fprintf(stderr, "Fresh board with random seed %lu\n", fast_getseed());
 226                 //board_print(b, stderr);
 227                 tree_load(u->t, b, color);
 228         }
 229
 230         /* XXX: We hope that the opponent didn't suddenly play
 231          * several moves in the row. */
 232         if (!is_resign(promote) && !tree_promote_at(u->t, b, promote)) {
 233                 if (UDEBUGL(2))
 234                         fprintf(stderr, "<cannot find node to promote>\n");
 235                 /* Reset tree */
 236                 tree_done(u->t);
 237                 u->t = tree_init(b, color);
 238         }
 239 }
 240
 241 /* Set in main thread in case the playouts should stop. */
 242 static volatile sig_atomic_t halt = 0;
 243
 244 static int
 245 uct_playouts(struct uct *u, struct board *b, enum stone color, struct tree *t)
 246 {
 247         int i, games = u->games;
 248         if (t->root->children)
 249                 games -= t->root->u.playouts / 1.5;
 250         /* else this is highly read-out but dead-end branch of opening book;
 251          * we need to start from scratch; XXX: Maybe actually base the readout
 252          * count based on number of playouts of best node? */
 253         for (i = 0; i < games; i++) {
 254                 int result = uct_playout(u, b, color, t);
 255                 if (result < 0) {
 256                         /* Tree descent has hit invalid move. */
 257                         continue;
 258                 }
 259
 260                 if (i > 0 && !(i % 10000)) {
 261                         progress_status(u, t, color, i);
 262                 }
 263
 264                 if (i > 0 && !(i % 500)) {
 265                         struct tree_node *best = u->policy->choose(u->policy, t->root, b, color);
 266                         if (best && best->u.playouts >= 5000 && best->u.value >= u->loss_threshold)
 267                                 break;
 268                 }
 269
 270                 if (halt) {
 271                         if (UDEBUGL(2))
 272                                 fprintf(stderr, "<halting early, %d games skipped>\n", games - i);
 273                         break;
 274                 }
 275         }
 276
 277         progress_status(u, t, color, i);
 278         if (UDEBUGL(3))
 279                 tree_dump(t, u->dumpthres);
 280         return i;
 281 }
 282
 283 static pthread_mutex_t finish_mutex = PTHREAD_MUTEX_INITIALIZER;
 284 static pthread_cond_t finish_cond = PTHREAD_COND_INITIALIZER;
 285 static volatile int finish_thread;
 286 static pthread_mutex_t finish_serializer = PTHREAD_MUTEX_INITIALIZER;
 287
 288 struct spawn_ctx {
 289         int tid;
 290         struct uct *u;
 291         struct board *b;
 292         enum stone color;
 293         struct tree *t;
 294         unsigned long seed;
 295         int games;
 296 };
 297
 298 static void *
 299 spawn_helper(void *ctx_)
 300 {
 301         struct spawn_ctx *ctx = ctx_;
 302         /* Setup */
 303         fast_srandom(ctx->seed);
 304         /* Run */
 305         ctx->games = uct_playouts(ctx->u, ctx->b, ctx->color, ctx->t);
 306         /* Finish */
 307         pthread_mutex_lock(&finish_serializer);
 308         pthread_mutex_lock(&finish_mutex);
 309         finish_thread = ctx->tid;
 310         pthread_cond_signal(&finish_cond);
 311         pthread_mutex_unlock(&finish_mutex);
 312         return ctx;
 313 }
 314
 315 static void
 316 uct_notify_play(struct engine *e, struct board *b, struct move *m)
 317 {
 318         prepare_move(e, b, stone_other(m->color), m->coord);
 319 }
 320
 321 static coord_t *
 322 uct_genmove(struct engine *e, struct board *b, enum stone color)
 323 {
 324         struct uct *u = e->data;
 325
 326         /* Seed the tree. */
 327         prepare_move(e, b, color, resign);
 328
 329         int played_games = 0;
 330         if (!u->threads) {
 331                 played_games = uct_playouts(u, b, color, u->t);
 332         } else {
 333                 pthread_t threads[u->threads];
 334                 int joined = 0;
 335                 halt = 0;
 336                 pthread_mutex_lock(&finish_mutex);
 337                 /* Spawn threads... */
 338                 for (int ti = 0; ti < u->threads; ti++) {
 339                         struct spawn_ctx *ctx = malloc(sizeof(*ctx));
 340                         ctx->u = u; ctx->b = b; ctx->color = color;
 341                         ctx->t = tree_copy(u->t); ctx->tid = ti;
 342                         ctx->seed = fast_random(65536) + ti;
 343                         pthread_create(&threads[ti], NULL, spawn_helper, ctx);
 344                         if (UDEBUGL(2))
 345                                 fprintf(stderr, "Spawned thread %d\n", ti);
 346                 }
 347                 /* ...and collect them back: */
 348                 while (joined < u->threads) {
 349                         /* Wait for some thread to finish... */
 350                         pthread_cond_wait(&finish_cond, &finish_mutex);
 351                         /* ...and gather its remnants. */
 352                         struct spawn_ctx *ctx;
 353                         pthread_join(threads[finish_thread], (void **) &ctx);
 354                         played_games += ctx->games;
 355                         joined++;
 356                         tree_merge(u->t, ctx->t);
 357                         tree_done(ctx->t);
 358                         free(ctx);
 359                         if (UDEBUGL(2))
 360                                 fprintf(stderr, "Joined thread %d\n", finish_thread);
 361                         /* Do not get stalled by slow threads. */
 362                         if (joined >= u->threads / 2)
 363                                 halt = 1;
 364                         pthread_mutex_unlock(&finish_serializer);
 365                 }
 366                 pthread_mutex_unlock(&finish_mutex);
 367         }
 368
 369         if (UDEBUGL(2))
 370                 tree_dump(u->t, u->dumpthres);
 371
 372         struct tree_node *best = u->policy->choose(u->policy, u->t->root, b, color);
 373         if (!best) {
 374                 tree_done(u->t); u->t = NULL;
 375                 return coord_copy(pass);
 376         }
 377         if (UDEBUGL(0))
 378                 progress_status(u, u->t, color, played_games);
 379         if (UDEBUGL(1))
 380                 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);
 381         if (best->u.value < u->resign_ratio && !is_pass(best->coord)) {
 382                 tree_done(u->t); u->t = NULL;
 383                 return coord_copy(resign);
 384         }
 385         tree_promote_node(u->t, best);
 386         return coord_copy(best->coord);
 387 }
 388
 389 bool
 390 uct_genbook(struct engine *e, struct board *b, enum stone color)
 391 {
 392         struct uct *u = e->data;
 393         u->t = tree_init(b, color);
 394         tree_load(u->t, b, color);
 395
 396         int i;
 397         for (i = 0; i < u->games; i++) {
 398                 int result = uct_playout(u, b, color, u->t);
 399                 if (result < 0) {
 400                         /* Tree descent has hit invalid move. */
 401                         continue;
 402                 }
 403
 404                 if (i > 0 && !(i % 10000)) {
 405                         progress_status(u, u->t, color, i);
 406                 }
 407         }
 408         progress_status(u, u->t, color, i);
 409
 410         tree_save(u->t, b, u->games / 100);
 411
 412         tree_done(u->t);
 413
 414         return true;
 415 }
 416
 417 void
 418 uct_dumpbook(struct engine *e, struct board *b, enum stone color)
 419 {
 420         struct uct *u = e->data;
 421         u->t = tree_init(b, color);
 422         tree_load(u->t, b, color);
 423         tree_dump(u->t, 0);
 424         tree_done(u->t);
 425 }
 426
 427
 428 struct uct *
 429 uct_state_init(char *arg)
 430 {
 431         struct uct *u = calloc(1, sizeof(struct uct));
 432
 433         u->debug_level = 1;
 434         u->games = MC_GAMES;
 435         u->gamelen = MC_GAMELEN;
 436         u->expand_p = 2;
 437         u->dumpthres = 1000;
 438         u->playout_amaf = false;
 439
 440         if (arg) {
 441                 char *optspec, *next = arg;
 442                 while (*next) {
 443                         optspec = next;
 444                         next += strcspn(next, ",");
 445                         if (*next) { *next++ = 0; } else { *next = 0; }
 446
 447                         char *optname = optspec;
 448                         char *optval = strchr(optspec, '=');
 449                         if (optval) *optval++ = 0;
 450
 451                         if (!strcasecmp(optname, "debug")) {
 452                                 if (optval)
 453                                         u->debug_level = atoi(optval);
 454                                 else
 455                                         u->debug_level++;
 456                         } else if (!strcasecmp(optname, "games") && optval) {
 457                                 u->games = atoi(optval);
 458                         } else if (!strcasecmp(optname, "gamelen") && optval) {
 459                                 u->gamelen = atoi(optval);
 460                         } else if (!strcasecmp(optname, "expand_p") && optval) {
 461                                 u->expand_p = atoi(optval);
 462                         } else if (!strcasecmp(optname, "radar_d") && optval) {
 463                                 /* For 19x19, it is good idea to set this to 3. */
 464                                 u->radar_d = atoi(optval);
 465                         } else if (!strcasecmp(optname, "dumpthres") && optval) {
 466                                 u->dumpthres = atoi(optval);
 467                         } else if (!strcasecmp(optname, "playout_amaf")) {
 468                                 /* Whether to include random playout moves in
 469                                  * AMAF as well. (Otherwise, only tree moves
 470                                  * are included in AMAF. Of course makes sense
 471                                  * only in connection with an AMAF policy.) */
 472                                 /* with-without: 55.5% (+-4.1) */
 473                                 if (optval && *optval == '0')
 474                                         u->playout_amaf = false;
 475                                 else
 476                                         u->playout_amaf = true;
 477                         } else if (!strcasecmp(optname, "policy") && optval) {
 478                                 char *policyarg = strchr(optval, ':');
 479                                 if (policyarg)
 480                                         *policyarg++ = 0;
 481                                 if (!strcasecmp(optval, "ucb1")) {
 482                                         u->policy = policy_ucb1_init(u, policyarg);
 483                                 } else if (!strcasecmp(optval, "ucb1tuned")) {
 484                                         u->policy = policy_ucb1tuned_init(u, policyarg);
 485                                 } else if (!strcasecmp(optval, "ucb1amaf")) {
 486                                         u->policy = policy_ucb1amaf_init(u, policyarg);
 487                                 } else {
 488                                         fprintf(stderr, "UCT: Invalid tree policy %s\n", optval);
 489                                 }
 490                         } else if (!strcasecmp(optname, "playout") && optval) {
 491                                 char *playoutarg = strchr(optval, ':');
 492                                 if (playoutarg)
 493                                         *playoutarg++ = 0;
 494                                 if (!strcasecmp(optval, "moggy")) {
 495                                         u->playout = playout_moggy_init(playoutarg);
 496                                 } else if (!strcasecmp(optval, "light")) {
 497                                         u->playout = playout_light_init(playoutarg);
 498                                 } else {
 499                                         fprintf(stderr, "UCT: Invalid playout policy %s\n", optval);
 500                                 }
 501                         } else if (!strcasecmp(optname, "threads") && optval) {
 502                                 u->threads = atoi(optval);
 503                         } else if (!strcasecmp(optname, "force_seed") && optval) {
 504                                 u->force_seed = atoi(optval);
 505                         } else {
 506                                 fprintf(stderr, "uct: Invalid engine argument %s or missing value\n", optname);
 507                         }
 508                 }
 509         }
 510
 511         u->resign_ratio = 0.2; /* Resign when most games are lost. */
 512         u->loss_threshold = 0.85; /* Stop reading if after at least 5000 playouts this is best value. */
 513         if (!u->policy)
 514                 u->policy = policy_ucb1amaf_init(u, NULL);
 515
 516         if (!u->playout)
 517                 u->playout = playout_moggy_init(NULL);
 518         u->playout->debug_level = u->debug_level;
 519
 520         return u;
 521 }
 522
 523
 524 struct engine *
 525 engine_uct_init(char *arg)
 526 {
 527         struct uct *u = uct_state_init(arg);
 528         struct engine *e = calloc(1, sizeof(struct engine));
 529         e->name = "UCT Engine";
 530         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).";
 531         e->genmove = uct_genmove;
 532         e->notify_play = uct_notify_play;
 533         e->data = u;
 534
 535         return e;
 536 }