diff options
| author | tslil clingman <tslil@posteo.de> | 2023-01-15 21:31:00 +0100 |
|---|---|---|
| committer | tslil <tslil@posteo.de> | 2026-08-28 19:37:41 +0100 |
| commit | 0223a9bec5535fced1a7698b55fd42155d9b0446 (patch) | |
| tree | e7a980454e65d88b56194eed733cabec29ef51b5 /include/nn1986.c | |
| parent | ee216c008a188a9436fedb85c70ee5d1719733b1 (diff) | |
switch to explicit game state & important bug fix & clang format
Previously the code base assumed that there was a single, global game
state which was the implicit target of all actions taken. Looking
ahead at architectural improvements, this has now been (almost
entirely) made explicit and functions take tak_state_p where
necessary (and also where unnecessary).
Two important fixes to actions.c were made:
- Previously when generating the possible stack moves, stack height
overflows (> 15) were not taken into account and this resulted in the
tree search corrupting the board state. Now action search does not
list all legal actions, rather the subset of these encodeable by the
implementation.
- The check for crushing on a stack move was incorrect (too strict),
and this resulted in many legitimate moves being igonored.
Finally, in other changes, weights have also been improved by training
all games instead of some subset for chosen players, and clang-format
was run on the codebase.
Diffstat (limited to 'include/nn1986.c')
| -rw-r--r-- | include/nn1986.c | 26 |
1 files changed, 13 insertions, 13 deletions
diff --git a/include/nn1986.c b/include/nn1986.c index b9c551c..117a034 100644 --- a/include/nn1986.c +++ b/include/nn1986.c @@ -28,22 +28,22 @@ static float dense1[DENSE_NUM]; static float output[2]; #define RELU(x) ((x) = ((x) < 0) ? 0 : (x)) -float nn1986_evaluate_black_win(void) { +float nn1986_evaluate_black_win(tak_state_p state) { /* --------------- * * Populate input * * --------------- */ - for (unsigned int y = 0; y < board_size; y++) { - for (unsigned int x = 0; x < board_size; x++) { - const unsigned int loc = x + y * board_size; - const unsigned int count = COUNT_AT(loc); + for (unsigned int y = 0; y < state->board_size; y++) { + for (unsigned int x = 0; x < state->board_size; x++) { + const unsigned int loc = x + y * state->board_size; + const unsigned int count = COUNT_AT(state, loc); float lookup = 0; if (count > 0) { - if (STONE_AT(loc) == STONE_STANDING) { - lookup = (colours[loc] & 1) ? +0.25 : -0.25; - } else if (STONE_AT(loc) == STONE_CAPSTONE) { - lookup = (colours[loc] & 1) ? +1.00 : -1.00; + if (STONE_AT(state, loc) == STONE_STANDING) { + lookup = (state->colours[loc] & 1) ? +0.25 : -0.25; + } else if (STONE_AT(state, loc) == STONE_CAPSTONE) { + lookup = (state->colours[loc] & 1) ? +1.00 : -1.00; } else { - lookup = (colours[loc] & 1) ? +0.50 : -0.50; + lookup = (state->colours[loc] & 1) ? +0.50 : -0.50; } } cur_board[3 + loc] = lookup; @@ -53,9 +53,9 @@ float nn1986_evaluate_black_win(void) { * Convolution layer * * ------------------ */ // Add input of flat counts and ply parity - cur_board[0] = (ply & 1) ? 1 : -1; - cur_board[1] = (float)(white_count & 127) / 21.0; - cur_board[2] = (float)(black_count & 127) / 21.0; + cur_board[0] = (state->ply & 1) ? 1 : -1; + cur_board[1] = (float)(state->white_count & 127) / 21.0; + cur_board[2] = (float)(state->black_count & 127) / 21.0; /* ------------------ * * First dense layer * * ------------------ */ |
