Hugging Face's logo

thomas-schweich
/
pawn-base

Other
Safetensors
PyTorch
Rust
English
pawn
chess
transformer
world-model
causal-lm
next-token-prediction
representation-learning
Eval Results (legacy)
Model card Files
xet
 Community
thomas-schweich commited on
Commit
4ba32d9
·
verified ·
1 Parent(s): 89e81ca

Add eval results (probes + diagnostics)

Browse files
Files changed (1) hide show
  1. eval_results.json +170 -0
eval_results.json ADDED
@@ -0,0 +1,170 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "probes": {
3
+ "piece_type": {
4
+ "layer_7": {
5
+ "accuracy": 0.8972528332996329,
6
+ "loss": 0.462514906460532,
7
+ "best_accuracy": 0.8972528332996329,
8
+ "n_train": 485944,
9
+ "n_val": 120043
10
+ }
11
+ },
12
+ "side_to_move": {
13
+ "layer_7": {
14
+ "accuracy": 0.999999999883316,
15
+ "loss": 4.451842106768326e-08,
16
+ "best_accuracy": 0.999999999883316,
17
+ "n_train": 485944,
18
+ "n_val": 120043
19
+ }
20
+ },
21
+ "is_check": {
22
+ "layer_7": {
23
+ "accuracy": 0.9391384753419195,
24
+ "loss": 0.23080479027553283,
25
+ "best_accuracy": 0.942370650479162,
26
+ "n_train": 485944,
27
+ "n_val": 120043
28
+ }
29
+ },
30
+ "castling_rights": {
31
+ "layer_7": {
32
+ "accuracy": 0.9661996117109328,
33
+ "loss": 0.09139342949068821,
34
+ "best_accuracy": 0.9664266137326708,
35
+ "n_train": 485944,
36
+ "n_val": 120043
37
+ }
38
+ },
39
+ "ep_square": {
40
+ "layer_7": {
41
+ "accuracy": 0.9960014327032223,
42
+ "loss": 0.07782827074727709,
43
+ "best_accuracy": 0.9971426904192073,
44
+ "n_train": 485944,
45
+ "n_val": 120043
46
+ }
47
+ },
48
+ "material_count": {
49
+ "layer_7": {
50
+ "accuracy": 0.7550588543344782,
51
+ "loss": 0.7828635349641946,
52
+ "best_accuracy": 0.8613100342512295,
53
+ "n_train": 485944,
54
+ "n_val": 120043,
55
+ "mae": 6.08153298697524
56
+ }
57
+ },
58
+ "legal_move_count": {
59
+ "layer_7": {
60
+ "accuracy": 0.379081750813699,
61
+ "loss": 83.04724725417708,
62
+ "best_accuracy": 0.379081750813699,
63
+ "n_train": 485944,
64
+ "n_val": 120043,
65
+ "mae": 6.84791993114902
66
+ }
67
+ },
68
+ "halfmove_clock": {
69
+ "layer_7": {
70
+ "accuracy": 0.09443258406520243,
71
+ "loss": 36.30150203530533,
72
+ "best_accuracy": 0.11757128595164827,
73
+ "n_train": 485944,
74
+ "n_val": 120043,
75
+ "mae": 4.078397593157374
76
+ }
77
+ },
78
+ "game_phase": {
79
+ "layer_7": {
80
+ "accuracy": 0.9041093607915214,
81
+ "loss": 0.25151603218794977,
82
+ "best_accuracy": 0.907024982682253,
83
+ "n_train": 485944,
84
+ "n_val": 120043
85
+ }
86
+ }
87
+ },
88
+ "diagnostics": {
89
+ "in_check": {
90
+ "n_positions": 1000,
91
+ "mean_legal_rate": 0.9768800000000001,
92
+ "std_legal_rate": 0.06257527946401838,
93
+ "mean_pad_prob": 0.00044610700723531917,
94
+ "mean_entropy": 1.378539781215135,
95
+ "std_entropy": 0.4953501412320156
96
+ },
97
+ "double_check": {
98
+ "n_positions": 71,
99
+ "mean_legal_rate": 0.9123943661971831,
100
+ "std_legal_rate": 0.16090820398321937,
101
+ "mean_pad_prob": 0.0003617955171410851,
102
+ "mean_entropy": 1.1824169549828685,
103
+ "std_entropy": 0.4335250151114159
104
+ },
105
+ "pin_restricts": {
106
+ "n_positions": 1000,
107
+ "mean_legal_rate": 0.97175,
108
+ "std_legal_rate": 0.05368647408798607,
109
+ "mean_pad_prob": 1.9187790060342758e-05,
110
+ "mean_entropy": 3.195623012639582,
111
+ "std_entropy": 0.5958539652673973
112
+ },
113
+ "ep_available": {
114
+ "n_positions": 940,
115
+ "mean_legal_rate": 0.9922553191489362,
116
+ "std_legal_rate": 0.02910031665039855,
117
+ "mean_pad_prob": 1.1382038407927192e-06,
118
+ "mean_entropy": 3.4613142376250408,
119
+ "std_entropy": 0.325010232064062
120
+ },
121
+ "castle_legal_k": {
122
+ "n_positions": 1000,
123
+ "mean_legal_rate": 0.9974100000000001,
124
+ "std_legal_rate": 0.005813080078581411,
125
+ "mean_pad_prob": 3.8981330552659286e-07,
126
+ "mean_entropy": 3.5879413638114928,
127
+ "std_entropy": 0.17921419256677748
128
+ },
129
+ "castle_legal_q": {
130
+ "n_positions": 1000,
131
+ "mean_legal_rate": 0.9958800000000001,
132
+ "std_legal_rate": 0.007591152745136939,
133
+ "mean_pad_prob": 4.083307937790548e-07,
134
+ "mean_entropy": 3.6458762888908387,
135
+ "std_entropy": 0.18727308169032228
136
+ },
137
+ "castle_blocked_check": {
138
+ "n_positions": 892,
139
+ "mean_legal_rate": 0.9937107623318385,
140
+ "std_legal_rate": 0.01466430707907582,
141
+ "mean_pad_prob": 0.00044056765007829507,
142
+ "mean_entropy": 1.2424753979662841,
143
+ "std_entropy": 0.5342524588211677
144
+ },
145
+ "promotion_available": {
146
+ "n_positions": 1000,
147
+ "mean_legal_rate": 0.9940599999999999,
148
+ "std_legal_rate": 0.011735263098882794,
149
+ "mean_pad_prob": 1.245503282853887e-06,
150
+ "mean_entropy": 3.2927837179899218,
151
+ "std_entropy": 0.46145118335448565
152
+ },
153
+ "checkmate": {
154
+ "n_positions": 276,
155
+ "mean_legal_rate": 0.0,
156
+ "std_legal_rate": 0.0,
157
+ "mean_pad_prob": 0.9116162136483911,
158
+ "mean_entropy": 0.296485440773498,
159
+ "std_entropy": 0.3978962669047925
160
+ },
161
+ "stalemate": {
162
+ "n_positions": 41,
163
+ "mean_legal_rate": 0.0,
164
+ "std_legal_rate": 0.0,
165
+ "mean_pad_prob": 0.8422023932529022,
166
+ "mean_entropy": 0.30673794891889683,
167
+ "std_entropy": 0.46605586628775386
168
+ }
169
+ }
170
+ }