-
-
Notifications
You must be signed in to change notification settings - Fork 18
/
halfka_v2.py
106 lines (86 loc) · 3.9 KB
/
halfka_v2.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
import chess
import torch
import feature_block
from collections import OrderedDict
from feature_block import *
import variant
NUM_SQ = variant.SQUARES
NUM_KSQ = variant.KING_SQUARES
NUM_PT_REAL = variant.PIECES - (NUM_KSQ != 1)
NUM_PT_VIRTUAL = variant.PIECES
NUM_PLANES_REAL = NUM_SQ * NUM_PT_REAL + (NUM_PT_REAL - (NUM_KSQ != 1)) * variant.POCKETS
NUM_PLANES_VIRTUAL = NUM_SQ * NUM_PT_VIRTUAL + (NUM_PT_REAL - (NUM_KSQ != 1)) * variant.POCKETS
NUM_INPUTS = NUM_PLANES_REAL * NUM_KSQ
def orient(is_white_pov: bool, sq: int):
return sq % variant.FILES + (variant.RANKS - 1 - (sq // variant.FILES)) * variant.FILES if not is_white_pov else sq
def halfka_idx(is_white_pov: bool, king_sq: int, sq: int, piece_type: int, color: bool):
p_idx = (piece_type - 1) * 2 + (color != is_white_pov)
if NUM_PT_REAL % 2 and p_idx == NUM_PT_REAL:
# merge kings into one plane
p_idx -= 1
return orient(is_white_pov, sq) + p_idx * NUM_SQ + king_sq * NUM_PLANES_REAL
def halfka_hand_idx(is_white_pov: bool, king_sq: int, handCount: int, piece_type: int, color: bool):
p_idx = (piece_type - 1) * 2 + (color != is_white_pov)
return handCount + p_idx * variant.POCKETS + NUM_SQ * NUM_PT_REAL + king_sq * NUM_PLANES_REAL
def map_king(sq: int):
# palace squares for Xiangi/Janggi
if NUM_KSQ == 9 and NUM_KSQ != NUM_SQ:
if sq > variant.FILES * ((variant.RANKS + 1) // 2):
# in order to allow unambiguously detecting opposing kings, just return value out of range
return sq
# map accessible king squares skipping the gaps
return (sq - 6 * (sq // variant.FILES) - 3) % NUM_KSQ
return sq % NUM_KSQ
def halfka_psqts():
values = [0] * (NUM_PLANES_REAL * NUM_KSQ)
for ksq in range(NUM_KSQ):
for s in range(NUM_SQ):
for pt, val in variant.PIECE_VALUES.items():
idxw = halfka_idx(True, ksq, s, pt, chess.WHITE)
idxb = halfka_idx(True, ksq, s, pt, chess.BLACK)
values[idxw] = val
values[idxb] = -val
for i in range(variant.POCKETS):
for pt, val in variant.PIECE_VALUES.items():
idxw = halfka_hand_idx(True, ksq, i, pt, chess.WHITE)
idxb = halfka_hand_idx(True, ksq, i, pt, chess.BLACK)
values[idxw] = val
values[idxb] = -val
return values
class Features(FeatureBlock):
def __init__(self):
super(Features, self).__init__('HalfKAv2', 0x5f234cb8, OrderedDict([('HalfKAv2', NUM_PLANES_REAL * NUM_KSQ)]))
def get_active_features(self, board: chess.Board):
def piece_features(turn):
indices = torch.zeros(NUM_PLANES_REAL * NUM_KSQ)
for sq, p in board.piece_map().items():
indices[halfka_idx(turn, orient(turn, board.king(turn)), sq, p)] = 1.0
return indices
return (piece_features(chess.WHITE), piece_features(chess.BLACK))
def get_initial_psqt_features(self):
return halfka_psqts()
class FactorizedFeatures(FeatureBlock):
def __init__(self):
super(FactorizedFeatures, self).__init__('HalfKAv2^', 0x5f234cb8, OrderedDict([('HalfKAv2', NUM_PLANES_REAL * NUM_KSQ), ('A', NUM_PLANES_VIRTUAL)]))
def get_active_features(self, board: chess.Board):
raise Exception('Not supported yet, you must use the c++ data loader for factorizer support during training')
def get_feature_factors(self, idx):
if idx >= self.num_real_features:
raise Exception('Feature must be real')
a_idx = idx % NUM_PLANES_REAL
k_idx = idx // NUM_PLANES_REAL
if NUM_PT_VIRTUAL != NUM_PT_REAL:
if a_idx // NUM_SQ == NUM_PT_REAL - 1 and k_idx != map_king(a_idx % NUM_SQ):
# is king piece, but not ours
a_idx += NUM_SQ
elif a_idx >= NUM_SQ * NUM_PT_REAL:
# pockets
a_idx += NUM_SQ
return [idx, self.get_factor_base_feature('A') + a_idx]
def get_initial_psqt_features(self):
return halfka_psqts() + [0] * NUM_PLANES_VIRTUAL
'''
This is used by the features module for discovery of feature blocks.
'''
def get_feature_block_clss():
return [Features, FactorizedFeatures]