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