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 (passes >= 2) {
 185                 float score = board_official_score(&b2);
 186                 result = (player_color == S_BLACK) ? score < 0 : score > 0;
 187
 188                 if (UDEBUGL(5))
 189                         fprintf(stderr, "[%d..%d] %s p-p scoring playout result %d (W %f)\n",
 190                                 player_color, node_color, coord2sstr(n->coord, t->board), result, score);
 191                 if (UDEBUGL(6))
 192                         board_print(&b2, stderr);
 193
 194         } else { assert(tree_leaf_node(n));
 195                 result = uct_leaf_node(u, &b2, player_color, amaf, t, n, node_color, spaces);
 196         }
 197
 198         assert(n == t->root || n->parent);
 199         if (result >= 0)
 200                 u->policy->update(u->policy, t, n, node_color, player_color, amaf, result);
 201
 202 end:
 203         if (amaf) {
 204                 free(amaf->map - 1);
 205                 free(amaf);
 206         }
 207         board_done_noalloc(&b2);
 208         return result;
 209 }
 210
 211 static void
 212 prepare_move(struct engine *e, struct board *b, enum stone color, coord_t promote)
 213 {
 214         struct uct *u = e->data;
 215
 216         if (!b->moves && u->t) {
 217                 /* Stale state from last game */
 218                 tree_done(u->t);
 219                 u->t = NULL;
 220         }
 221
 222         if (!u->t) {
 223                 u->t = tree_init(b, color);
 224                 if (u->force_seed)
 225                         fast_srandom(u->force_seed);
 226                 if (UDEBUGL(0))
 227                         fprintf(stderr, "Fresh board with random seed %lu\n", fast_getseed());
 228                 //board_print(b, stderr);
 229                 tree_load(u->t, b, color);
 230         }
 231
 232         /* XXX: We hope that the opponent didn't suddenly play
 233          * several moves in the row. */
 234         if (!is_resign(promote) && !tree_promote_at(u->t, b, promote)) {
 235                 if (UDEBUGL(2))
 236                         fprintf(stderr, "<cannot find node to promote>\n");
 237                 /* Reset tree */
 238                 tree_done(u->t);
 239                 u->t = tree_init(b, color);
 240         }
 241 }
 242
 243 /* Set in main thread in case the playouts should stop. */
 244 static volatile sig_atomic_t halt = 0;
 245
 246 static int
 247 uct_playouts(struct uct *u, struct board *b, enum stone color, struct tree *t)
 248 {
 249         int i, games = u->games;
 250         if (t->root->children)
 251                 games -= t->root->u.playouts / 1.5;
 252         /* else this is highly read-out but dead-end branch of opening book;
 253          * we need to start from scratch; XXX: Maybe actually base the readout
 254          * count based on number of playouts of best node? */
 255         for (i = 0; i < games; i++) {
 256                 int result = uct_playout(u, b, color, t);
 257                 if (result < 0) {
 258                         /* Tree descent has hit invalid move. */
 259                         continue;
 260                 }
 261
 262                 if (i > 0 && !(i % 10000)) {
 263                         progress_status(u, t, color, i);
 264                 }
 265
 266                 if (i > 0 && !(i % 500)) {
 267                         struct tree_node *best = u->policy->choose(u->policy, t->root, b, color);
 268                         if (best && best->u.playouts >= 5000 && best->u.value >= u->loss_threshold)
 269                                 break;
 270                 }
 271
 272                 if (halt) {
 273                         if (UDEBUGL(2))
 274                                 fprintf(stderr, "<halting early, %d games skipped>\n", games - i);
 275                         break;
 276                 }
 277         }
 278
 279         progress_status(u, t, color, i);
 280         if (UDEBUGL(3))
 281                 tree_dump(t, u->dumpthres);
 282         return i;
 283 }
 284
 285 static pthread_mutex_t finish_mutex = PTHREAD_MUTEX_INITIALIZER;
 286 static pthread_cond_t finish_cond = PTHREAD_COND_INITIALIZER;
 287 static volatile int finish_thread;
 288 static pthread_mutex_t finish_serializer = PTHREAD_MUTEX_INITIALIZER;
 289
 290 struct spawn_ctx {
 291         int tid;
 292         struct uct *u;
 293         struct board *b;
 294         enum stone color;
 295         struct tree *t;
 296         unsigned long seed;
 297         int games;
 298 };
 299
 300 static void *
 301 spawn_helper(void *ctx_)
 302 {
 303         struct spawn_ctx *ctx = ctx_;
 304         /* Setup */
 305         fast_srandom(ctx->seed);
 306         /* Run */
 307         ctx->games = uct_playouts(ctx->u, ctx->b, ctx->color, ctx->t);
 308         /* Finish */
 309         pthread_mutex_lock(&finish_serializer);
 310         pthread_mutex_lock(&finish_mutex);
 311         finish_thread = ctx->tid;
 312         pthread_cond_signal(&finish_cond);
 313         pthread_mutex_unlock(&finish_mutex);
 314         return ctx;
 315 }
 316
 317 static void
 318 uct_notify_play(struct engine *e, struct board *b, struct move *m)
 319 {
 320         prepare_move(e, b, stone_other(m->color), m->coord);
 321 }
 322
 323 static coord_t *
 324 uct_genmove(struct engine *e, struct board *b, enum stone color)
 325 {
 326         struct uct *u = e->data;
 327
 328         /* Seed the tree. */
 329         prepare_move(e, b, color, resign);
 330
 331         int played_games = 0;
 332         if (!u->threads) {
 333                 played_games = uct_playouts(u, b, color, u->t);
 334         } else {
 335                 pthread_t threads[u->threads];
 336                 int joined = 0;
 337                 halt = 0;
 338                 pthread_mutex_lock(&finish_mutex);
 339                 /* Spawn threads... */
 340                 for (int ti = 0; ti < u->threads; ti++) {
 341                         struct spawn_ctx *ctx = malloc(sizeof(*ctx));
 342                         ctx->u = u; ctx->b = b; ctx->color = color;
 343                         ctx->t = tree_copy(u->t); ctx->tid = ti;
 344                         ctx->seed = fast_random(65536) + ti;
 345                         pthread_create(&threads[ti], NULL, spawn_helper, ctx);
 346                         if (UDEBUGL(2))
 347                                 fprintf(stderr, "Spawned thread %d\n", ti);
 348                 }
 349                 /* ...and collect them back: */
 350                 while (joined < u->threads) {
 351                         /* Wait for some thread to finish... */
 352                         pthread_cond_wait(&finish_cond, &finish_mutex);
 353                         /* ...and gather its remnants. */
 354                         struct spawn_ctx *ctx;
 355                         pthread_join(threads[finish_thread], (void **) &ctx);
 356                         played_games += ctx->games;
 357                         joined++;
 358                         tree_merge(u->t, ctx->t);
 359                         tree_done(ctx->t);
 360                         free(ctx);
 361                         if (UDEBUGL(2))
 362                                 fprintf(stderr, "Joined thread %d\n", finish_thread);
 363                         /* Do not get stalled by slow threads. */
 364                         if (joined >= u->threads / 2)
 365                                 halt = 1;
 366                         pthread_mutex_unlock(&finish_serializer);
 367                 }
 368                 pthread_mutex_unlock(&finish_mutex);
 369         }
 370
 371         if (UDEBUGL(2))
 372                 tree_dump(u->t, u->dumpthres);
 373
 374         struct tree_node *best = u->policy->choose(u->policy, u->t->root, b, color);
 375         if (!best) {
 376                 tree_done(u->t); u->t = NULL;
 377                 return coord_copy(pass);
 378         }
 379         if (UDEBUGL(0))
 380                 progress_status(u, u->t, color, played_games);
 381         if (UDEBUGL(1))
 382                 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);
 383         if (best->u.value < u->resign_ratio && !is_pass(best->coord)) {
 384                 tree_done(u->t); u->t = NULL;
 385                 return coord_copy(resign);
 386         }
 387         tree_promote_node(u->t, best);
 388         return coord_copy(best->coord);
 389 }
 390
 391 bool
 392 uct_genbook(struct engine *e, struct board *b, enum stone color)
 393 {
 394         struct uct *u = e->data;
 395         u->t = tree_init(b, color);
 396         tree_load(u->t, b, color);
 397
 398         int i;
 399         for (i = 0; i < u->games; i++) {
 400                 int result = uct_playout(u, b, color, u->t);
 401                 if (result < 0) {
 402                         /* Tree descent has hit invalid move. */
 403                         continue;
 404                 }
 405
 406                 if (i > 0 && !(i % 10000)) {
 407                         progress_status(u, u->t, color, i);
 408                 }
 409         }
 410         progress_status(u, u->t, color, i);
 411
 412         tree_save(u->t, b, u->games / 100);
 413
 414         tree_done(u->t);
 415
 416         return true;
 417 }
 418
 419 void
 420 uct_dumpbook(struct engine *e, struct board *b, enum stone color)
 421 {
 422         struct uct *u = e->data;
 423         u->t = tree_init(b, color);
 424         tree_load(u->t, b, color);
 425         tree_dump(u->t, 0);
 426         tree_done(u->t);
 427 }
 428
 429
 430 struct uct *
 431 uct_state_init(char *arg)
 432 {
 433         struct uct *u = calloc(1, sizeof(struct uct));
 434
 435         u->debug_level = 1;
 436         u->games = MC_GAMES;
 437         u->gamelen = MC_GAMELEN;
 438         u->expand_p = 2;
 439         u->dumpthres = 1000;
 440         u->playout_amaf = false;
 441
 442         if (arg) {
 443                 char *optspec, *next = arg;
 444                 while (*next) {
 445                         optspec = next;
 446                         next += strcspn(next, ",");
 447                         if (*next) { *next++ = 0; } else { *next = 0; }
 448
 449                         char *optname = optspec;
 450                         char *optval = strchr(optspec, '=');
 451                         if (optval) *optval++ = 0;
 452
 453                         if (!strcasecmp(optname, "debug")) {
 454                                 if (optval)
 455                                         u->debug_level = atoi(optval);
 456                                 else
 457                                         u->debug_level++;
 458                         } else if (!strcasecmp(optname, "games") && optval) {
 459                                 u->games = atoi(optval);
 460                         } else if (!strcasecmp(optname, "gamelen") && optval) {
 461                                 u->gamelen = atoi(optval);
 462                         } else if (!strcasecmp(optname, "expand_p") && optval) {
 463                                 u->expand_p = atoi(optval);
 464                         } else if (!strcasecmp(optname, "radar_d") && optval) {
 465                                 /* For 19x19, it is good idea to set this to 3. */
 466                                 u->radar_d = atoi(optval);
 467                         } else if (!strcasecmp(optname, "dumpthres") && optval) {
 468                                 u->dumpthres = atoi(optval);
 469                         } else if (!strcasecmp(optname, "playout_amaf")) {
 470                                 /* Whether to include random playout moves in
 471                                  * AMAF as well. (Otherwise, only tree moves
 472                                  * are included in AMAF. Of course makes sense
 473                                  * only in connection with an AMAF policy.) */
 474                                 /* with-without: 55.5% (+-4.1) */
 475                                 if (optval && *optval == '0')
 476                                         u->playout_amaf = false;
 477                                 else
 478                                         u->playout_amaf = true;
 479                         } else if (!strcasecmp(optname, "policy") && optval) {
 480                                 char *policyarg = strchr(optval, ':');
 481                                 if (policyarg)
 482                                         *policyarg++ = 0;
 483                                 if (!strcasecmp(optval, "ucb1")) {
 484                                         u->policy = policy_ucb1_init(u, policyarg);
 485                                 } else if (!strcasecmp(optval, "ucb1tuned")) {
 486                                         u->policy = policy_ucb1tuned_init(u, policyarg);
 487                                 } else if (!strcasecmp(optval, "ucb1amaf")) {
 488                                         u->policy = policy_ucb1amaf_init(u, policyarg);
 489                                 } else {
 490                                         fprintf(stderr, "UCT: Invalid tree policy %s\n", optval);
 491                                 }
 492                         } else if (!strcasecmp(optname, "playout") && optval) {
 493                                 char *playoutarg = strchr(optval, ':');
 494                                 if (playoutarg)
 495                                         *playoutarg++ = 0;
 496                                 if (!strcasecmp(optval, "moggy")) {
 497                                         u->playout = playout_moggy_init(playoutarg);
 498                                 } else if (!strcasecmp(optval, "light")) {
 499                                         u->playout = playout_light_init(playoutarg);
 500                                 } else {
 501                                         fprintf(stderr, "UCT: Invalid playout policy %s\n", optval);
 502                                 }
 503                         } else if (!strcasecmp(optname, "threads") && optval) {
 504                                 u->threads = atoi(optval);
 505                         } else if (!strcasecmp(optname, "force_seed") && optval) {
 506                                 u->force_seed = atoi(optval);
 507                         } else {
 508                                 fprintf(stderr, "uct: Invalid engine argument %s or missing value\n", optname);
 509                         }
 510                 }
 511         }
 512
 513         u->resign_ratio = 0.2; /* Resign when most games are lost. */
 514         u->loss_threshold = 0.85; /* Stop reading if after at least 5000 playouts this is best value. */
 515         if (!u->policy)
 516                 u->policy = policy_ucb1amaf_init(u, NULL);
 517
 518         if (!u->playout)
 519                 u->playout = playout_moggy_init(NULL);
 520         u->playout->debug_level = u->debug_level;
 521
 522         return u;
 523 }
 524
 525
 526 struct engine *
 527 engine_uct_init(char *arg)
 528 {
 529         struct uct *u = uct_state_init(arg);
 530         struct engine *e = calloc(1, sizeof(struct engine));
 531         e->name = "UCT Engine";
 532         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).";
 533         e->genmove = uct_genmove;
 534         e->notify_play = uct_notify_play;
 535         e->data = u;
 536
 537         return e;
 538 }