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         assert(node_color == t->root_color);
 128
 129         int result;
 130         int pass_limit = (board_size(&b2) - 2) * (board_size(&b2) - 2) / 2;
 131         int passes = is_pass(b->last_move.coord);
 132
 133         /* debug */
 134         int depth = 0;
 135         static char spaces[] = "\0                                                      ";
 136         /* /debug */
 137         if (UDEBUGL(8))
 138                 fprintf(stderr, "--- UCT walk with color %d\n", player_color);
 139
 140         while (!tree_leaf_node(n) && passes < 2) {
 141                 spaces[depth++] = ' '; spaces[depth] = 0;
 142
 143                 /* Parity is chosen already according to the child color, since
 144                  * it is applied to children. */
 145                 node_color = stone_other(node_color);
 146                 n = u->policy->descend(u->policy, t, n, (node_color == player_color ? 1 : -1), pass_limit);
 147
 148                 assert(n == t->root || n->parent);
 149                 if (UDEBUGL(7))
 150                         fprintf(stderr, "%s+-- UCT sent us to [%s:%d] %f\n",
 151                                 spaces, coord2sstr(n->coord, t->board), n->coord, n->u.value);
 152
 153                 if (amaf && n->coord >= -1 && !is_pass(n->coord)) {
 154                         if (amaf->map[n->coord] == S_NONE) {
 155                                 amaf->map[n->coord] = node_color;
 156                         } else {
 157                                 amaf_op(amaf->map[n->coord], +);
 158                         }
 159                 }
 160
 161                 struct move m = { n->coord, node_color };
 162                 int res = board_play(&b2, &m);
 163
 164                 if (res < 0 || (!is_pass(m.coord) && !group_at(&b2, m.coord)) /* suicide */
 165                     || b2.superko_violation) {
 166                         if (UDEBUGL(3)) {
 167                                 for (struct tree_node *ni = n; ni; ni = ni->parent)
 168                                         fprintf(stderr, "%s ", coord2sstr(ni->coord, t->board));
 169                                 fprintf(stderr, "deleting invalid %s node %d,%d res %d group %d spk %d\n",
 170                                         stone2str(node_color), coord_x(n->coord,b), coord_y(n->coord,b),
 171                                         res, group_at(&b2, m.coord), b2.superko_violation);
 172                         }
 173                         tree_delete_node(t, n);
 174                         result = -1;
 175                         goto end;
 176                 }
 177
 178                 if (is_pass(n->coord))
 179                         passes++;
 180                 else
 181                         passes = 0;
 182         }
 183
 184         if (tree_leaf_node(n)) {
 185                 result = uct_leaf_node(u, &b2, player_color, amaf, t, n, node_color, spaces);
 186
 187         } else { assert(passes >= 2);
 188
 189                 float score = board_official_score(&b2);
 190                 result = (player_color == S_BLACK) ? score < 0 : score > 0;
 191
 192                 if (UDEBUGL(5))
 193                         fprintf(stderr, "[%d..%d] %s p-p scoring playout result %d (W %f)\n",
 194                                 player_color, node_color, coord2sstr(n->coord, t->board), result, score);
 195                 if (UDEBUGL(6))
 196                         board_print(&b2, stderr);
 197         }
 198
 199         assert(n == t->root || n->parent);
 200         if (result >= 0)
 201                 u->policy->update(u->policy, t, n, node_color, player_color, amaf, result);
 202
 203 end:
 204         if (amaf) {
 205                 free(amaf->map - 1);
 206                 free(amaf);
 207         }
 208         board_done_noalloc(&b2);
 209         return result;
 210 }
 211
 212 static void
 213 prepare_move(struct engine *e, struct board *b, enum stone color, coord_t promote)
 214 {
 215         struct uct *u = e->data;
 216
 217         if (!b->moves && u->t) {
 218                 /* Stale state from last game */
 219                 tree_done(u->t);
 220                 u->t = NULL;
 221         }
 222
 223         if (!u->t) {
 224                 u->t = tree_init(b, color);
 225                 if (u->force_seed)
 226                         fast_srandom(u->force_seed);
 227                 if (UDEBUGL(0))
 228                         fprintf(stderr, "Fresh board with random seed %lu\n", fast_getseed());
 229                 //board_print(b, stderr);
 230                 tree_load(u->t, b, color);
 231         }
 232
 233         /* XXX: We hope that the opponent didn't suddenly play
 234          * several moves in the row. */
 235         if (!is_resign(promote) && !tree_promote_at(u->t, b, promote)) {
 236                 if (UDEBUGL(2))
 237                         fprintf(stderr, "<cannot find node to promote>\n");
 238                 /* Reset tree */
 239                 tree_done(u->t);
 240                 u->t = tree_init(b, color);
 241         }
 242 }
 243
 244 /* Set in main thread in case the playouts should stop. */
 245 static volatile sig_atomic_t halt = 0;
 246
 247 static int
 248 uct_playouts(struct uct *u, struct board *b, enum stone color, struct tree *t)
 249 {
 250         int i, games = u->games;
 251         if (t->root->children)
 252                 games -= t->root->u.playouts / 1.5;
 253         /* else this is highly read-out but dead-end branch of opening book;
 254          * we need to start from scratch; XXX: Maybe actually base the readout
 255          * count based on number of playouts of best node? */
 256         for (i = 0; i < games; i++) {
 257                 int result = uct_playout(u, b, color, t);
 258                 if (result < 0) {
 259                         /* Tree descent has hit invalid move. */
 260                         continue;
 261                 }
 262
 263                 if (i > 0 && !(i % 10000)) {
 264                         progress_status(u, t, color, i);
 265                 }
 266
 267                 if (i > 0 && !(i % 500)) {
 268                         struct tree_node *best = u->policy->choose(u->policy, t->root, b, color);
 269                         if (best && best->u.playouts >= 5000 && best->u.value >= u->loss_threshold)
 270                                 break;
 271                 }
 272
 273                 if (halt) {
 274                         if (UDEBUGL(2))
 275                                 fprintf(stderr, "<halting early, %d games skipped>\n", games - i);
 276                         break;
 277                 }
 278         }
 279
 280         progress_status(u, t, color, i);
 281         if (UDEBUGL(3))
 282                 tree_dump(t, u->dumpthres);
 283         return i;
 284 }
 285
 286 static pthread_mutex_t finish_mutex = PTHREAD_MUTEX_INITIALIZER;
 287 static pthread_cond_t finish_cond = PTHREAD_COND_INITIALIZER;
 288 static volatile int finish_thread;
 289 static pthread_mutex_t finish_serializer = PTHREAD_MUTEX_INITIALIZER;
 290
 291 struct spawn_ctx {
 292         int tid;
 293         struct uct *u;
 294         struct board *b;
 295         enum stone color;
 296         struct tree *t;
 297         unsigned long seed;
 298         int games;
 299 };
 300
 301 static void *
 302 spawn_helper(void *ctx_)
 303 {
 304         struct spawn_ctx *ctx = ctx_;
 305         /* Setup */
 306         fast_srandom(ctx->seed);
 307         /* Run */
 308         ctx->games = uct_playouts(ctx->u, ctx->b, ctx->color, ctx->t);
 309         /* Finish */
 310         pthread_mutex_lock(&finish_serializer);
 311         pthread_mutex_lock(&finish_mutex);
 312         finish_thread = ctx->tid;
 313         pthread_cond_signal(&finish_cond);
 314         pthread_mutex_unlock(&finish_mutex);
 315         return ctx;
 316 }
 317
 318 static void
 319 uct_notify_play(struct engine *e, struct board *b, struct move *m)
 320 {
 321         prepare_move(e, b, stone_other(m->color), m->coord);
 322 }
 323
 324 static coord_t *
 325 uct_genmove(struct engine *e, struct board *b, enum stone color)
 326 {
 327         struct uct *u = e->data;
 328
 329         /* Seed the tree. */
 330         prepare_move(e, b, color, resign);
 331
 332         int played_games = 0;
 333         if (!u->threads) {
 334                 played_games = uct_playouts(u, b, color, u->t);
 335         } else {
 336                 pthread_t threads[u->threads];
 337                 int joined = 0;
 338                 halt = 0;
 339                 pthread_mutex_lock(&finish_mutex);
 340                 /* Spawn threads... */
 341                 for (int ti = 0; ti < u->threads; ti++) {
 342                         struct spawn_ctx *ctx = malloc(sizeof(*ctx));
 343                         ctx->u = u; ctx->b = b; ctx->color = color;
 344                         ctx->t = tree_copy(u->t); ctx->tid = ti;
 345                         ctx->seed = fast_random(65536) + ti;
 346                         pthread_create(&threads[ti], NULL, spawn_helper, ctx);
 347                         if (UDEBUGL(2))
 348                                 fprintf(stderr, "Spawned thread %d\n", ti);
 349                 }
 350                 /* ...and collect them back: */
 351                 while (joined < u->threads) {
 352                         /* Wait for some thread to finish... */
 353                         pthread_cond_wait(&finish_cond, &finish_mutex);
 354                         /* ...and gather its remnants. */
 355                         struct spawn_ctx *ctx;
 356                         pthread_join(threads[finish_thread], (void **) &ctx);
 357                         played_games += ctx->games;
 358                         joined++;
 359                         tree_merge(u->t, ctx->t);
 360                         tree_done(ctx->t);
 361                         free(ctx);
 362                         if (UDEBUGL(2))
 363                                 fprintf(stderr, "Joined thread %d\n", finish_thread);
 364                         /* Do not get stalled by slow threads. */
 365                         if (joined >= u->threads / 2)
 366                                 halt = 1;
 367                         pthread_mutex_unlock(&finish_serializer);
 368                 }
 369                 pthread_mutex_unlock(&finish_mutex);
 370         }
 371
 372         if (UDEBUGL(2))
 373                 tree_dump(u->t, u->dumpthres);
 374
 375         struct tree_node *best = u->policy->choose(u->policy, u->t->root, b, color);
 376         if (!best) {
 377                 tree_done(u->t); u->t = NULL;
 378                 return coord_copy(pass);
 379         }
 380         if (UDEBUGL(0))
 381                 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);
 382         if (best->u.value < u->resign_ratio && !is_pass(best->coord)) {
 383                 tree_done(u->t); u->t = NULL;
 384                 return coord_copy(resign);
 385         }
 386         tree_promote_node(u->t, best);
 387         return coord_copy(best->coord);
 388 }
 389
 390 bool
 391 uct_genbook(struct engine *e, struct board *b, enum stone color)
 392 {
 393         struct uct *u = e->data;
 394         u->t = tree_init(b, color);
 395         tree_load(u->t, b, color);
 396
 397         int i;
 398         for (i = 0; i < u->games; i++) {
 399                 int result = uct_playout(u, b, color, u->t);
 400                 if (result < 0) {
 401                         /* Tree descent has hit invalid move. */
 402                         continue;
 403                 }
 404
 405                 if (i > 0 && !(i % 10000)) {
 406                         progress_status(u, u->t, color, i);
 407                 }
 408         }
 409         progress_status(u, u->t, color, i);
 410
 411         tree_save(u->t, b, u->games / 100);
 412
 413         tree_done(u->t);
 414
 415         return true;
 416 }
 417
 418 void
 419 uct_dumpbook(struct engine *e, struct board *b, enum stone color)
 420 {
 421         struct uct *u = e->data;
 422         u->t = tree_init(b, color);
 423         tree_load(u->t, b, color);
 424         tree_dump(u->t, 0);
 425         tree_done(u->t);
 426 }
 427
 428
 429 struct uct *
 430 uct_state_init(char *arg)
 431 {
 432         struct uct *u = calloc(1, sizeof(struct uct));
 433
 434         u->debug_level = 1;
 435         u->games = MC_GAMES;
 436         u->gamelen = MC_GAMELEN;
 437         u->expand_p = 2;
 438         u->dumpthres = 1000;
 439         u->playout_amaf = false;
 440
 441         if (arg) {
 442                 char *optspec, *next = arg;
 443                 while (*next) {
 444                         optspec = next;
 445                         next += strcspn(next, ",");
 446                         if (*next) { *next++ = 0; } else { *next = 0; }
 447
 448                         char *optname = optspec;
 449                         char *optval = strchr(optspec, '=');
 450                         if (optval) *optval++ = 0;
 451
 452                         if (!strcasecmp(optname, "debug")) {
 453                                 if (optval)
 454                                         u->debug_level = atoi(optval);
 455                                 else
 456                                         u->debug_level++;
 457                         } else if (!strcasecmp(optname, "games") && optval) {
 458                                 u->games = atoi(optval);
 459                         } else if (!strcasecmp(optname, "gamelen") && optval) {
 460                                 u->gamelen = atoi(optval);
 461                         } else if (!strcasecmp(optname, "expand_p") && optval) {
 462                                 u->expand_p = atoi(optval);
 463                         } else if (!strcasecmp(optname, "radar_d") && optval) {
 464                                 /* For 19x19, it is good idea to set this to 3. */
 465                                 u->radar_d = atoi(optval);
 466                         } else if (!strcasecmp(optname, "dumpthres") && optval) {
 467                                 u->dumpthres = atoi(optval);
 468                         } else if (!strcasecmp(optname, "playout_amaf")) {
 469                                 /* Whether to include random playout moves in
 470                                  * AMAF as well. (Otherwise, only tree moves
 471                                  * are included in AMAF. Of course makes sense
 472                                  * only in connection with an AMAF policy.) */
 473                                 /* with-without: 55.5% (+-4.1) */
 474                                 if (optval && *optval == '0')
 475                                         u->playout_amaf = false;
 476                                 else
 477                                         u->playout_amaf = true;
 478                         } else if (!strcasecmp(optname, "policy") && optval) {
 479                                 char *policyarg = strchr(optval, ':');
 480                                 if (policyarg)
 481                                         *policyarg++ = 0;
 482                                 if (!strcasecmp(optval, "ucb1")) {
 483                                         u->policy = policy_ucb1_init(u, policyarg);
 484                                 } else if (!strcasecmp(optval, "ucb1tuned")) {
 485                                         u->policy = policy_ucb1tuned_init(u, policyarg);
 486                                 } else if (!strcasecmp(optval, "ucb1amaf")) {
 487                                         u->policy = policy_ucb1amaf_init(u, policyarg);
 488                                 } else {
 489                                         fprintf(stderr, "UCT: Invalid tree policy %s\n", optval);
 490                                 }
 491                         } else if (!strcasecmp(optname, "playout") && optval) {
 492                                 char *playoutarg = strchr(optval, ':');
 493                                 if (playoutarg)
 494                                         *playoutarg++ = 0;
 495                                 if (!strcasecmp(optval, "moggy")) {
 496                                         u->playout = playout_moggy_init(playoutarg);
 497                                 } else if (!strcasecmp(optval, "light")) {
 498                                         u->playout = playout_light_init(playoutarg);
 499                                 } else {
 500                                         fprintf(stderr, "UCT: Invalid playout policy %s\n", optval);
 501                                 }
 502                         } else if (!strcasecmp(optname, "threads") && optval) {
 503                                 u->threads = atoi(optval);
 504                         } else if (!strcasecmp(optname, "force_seed") && optval) {
 505                                 u->force_seed = atoi(optval);
 506                         } else {
 507                                 fprintf(stderr, "uct: Invalid engine argument %s or missing value\n", optname);
 508                         }
 509                 }
 510         }
 511
 512         u->resign_ratio = 0.2; /* Resign when most games are lost. */
 513         u->loss_threshold = 0.85; /* Stop reading if after at least 5000 playouts this is best value. */
 514         if (!u->policy)
 515                 u->policy = policy_ucb1amaf_init(u, NULL);
 516
 517         if (!u->playout)
 518                 u->playout = playout_moggy_init(NULL);
 519         u->playout->debug_level = u->debug_level;
 520
 521         return u;
 522 }
 523
 524
 525 struct engine *
 526 engine_uct_init(char *arg)
 527 {
 528         struct uct *u = uct_state_init(arg);
 529         struct engine *e = calloc(1, sizeof(struct engine));
 530         e->name = "UCT Engine";
 531         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).";
 532         e->genmove = uct_genmove;
 533         e->notify_play = uct_notify_play;
 534         e->data = u;
 535
 536         return e;
 537 }