-
Notifications
You must be signed in to change notification settings - Fork 6
/
Evaluation.cpp
73 lines (70 loc) · 1.29 KB
/
Evaluation.cpp
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
#include "Evaluation.h"
RawInput::RawInput(Board &board)
{
for (int i = 0; i < BLSIZE; i++)
if (board[i] == C_B)
{
feature[0][i] = 1;
feature[1][i] = 0;
}
else if (board[i]==C_W)
{
feature[0][i] = 0;
feature[1][i] = 1;
}
else
{
feature[0][i] = 0;
feature[1][i] = 0;
}
}
RawOutput getEvaluation(Board board, int col, NN *network, bool use_transform, int lastmove)
{
#if 0
RawOutput output;
for (int i = 0; i < BLSIZE; i++)
output.p[i] = 1.0f / BLSIZE;
output.v = 0;
return output;
#else
int trans;
if (use_transform)
trans = rand() % 8;
else
trans = 0;
boardTransform(trans, board);
Network::NNPlanes input;
input.resize(2);
for (int i = 0; i < BLSIZE; i++)
{
if (board[i] == 0)
{
input[0][i] = 0;
input[1][i] = 0;
}
else if (board[i] == col)
{
input[0][i] = 1;
input[1][i] = 0;
}
else
{
input[0][i] = 0;
input[1][i] = 1;
}
}
auto ret = network->forward(input);
double sum_policy = 0.0;
//scale sum_policy to 1
for (int i = 0; i<BLSIZE; i++)
if (board[i]==0)
sum_policy += ret.first[i];
RawOutput output;
if (sum_policy > 1e-10)
for (int i = 0; i<BLSIZE; i++)
output.p[i] = (float)(ret.first[i] / sum_policy);
boardTransform(trans + 8, output.p);
output.v = ret.second * 2.0f - 1.0f;
return output;
#endif
}