KeithG33 commited on
Commit
07e6790
·
1 Parent(s): e66e6eb

try remote strat

Browse files
Files changed (1) hide show
  1. ChessBot-Dataset.py +117 -39
ChessBot-Dataset.py CHANGED
@@ -1,59 +1,137 @@
1
- from datasets import (
2
- GeneratorBasedBuilder,
3
- DatasetInfo,
4
- SplitGenerator,
5
- Split,
6
- DownloadManager,
7
- Features,
8
- Value,
9
- Array2D,
10
- )
11
- from adversarial_gym.chess_env import ChessEnv
12
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
13
  class ChessPGNDataset(GeneratorBasedBuilder):
14
- VERSION = "0.0.1"
 
 
 
 
 
 
 
 
 
 
 
15
 
16
  def _info(self):
17
  return DatasetInfo(
18
- description="Chess positions + moves + results, streamed from PGN shards",
19
  features=Features({
20
  "state": Array2D((8,8), dtype="int8"),
21
  "action": Value("int16"),
22
  "result": Value("int8"),
23
- })
24
  )
25
 
26
  def _split_generators(self, dl_manager: DownloadManager):
27
- from pathlib import Path
28
- # 1) discover your shards inside the repo:
29
- data_dir = Path(self.config.data_dir)
30
- train_shards = sorted(data_dir.joinpath("train").glob("*.pgn"))
31
- test_shards = sorted(data_dir.joinpath("test").glob("*.pgn"))
32
-
33
- # 2) resolve them through the DL manager:
34
- # if they’re URLs, this downloads; if local, returns the same paths.
35
- train_paths = dl_manager.download([str(p) for p in train_shards])
36
- test_paths = dl_manager.download([str(p) for p in test_shards])
37
-
38
  return [
39
- SplitGenerator(
40
- name=Split.TRAIN,
41
- gen_kwargs={"shards": train_paths},
42
- ),
43
- SplitGenerator(
44
- name=Split.TEST,
45
- gen_kwargs={"shards": test_paths},
46
- ),
47
  ]
 
48
  def _generate_examples(self, shards):
49
- import chess.pgn
50
- uid = 0
51
 
 
52
  for path in sorted(shards):
53
- with open(path, "r") as f:
54
  while (game := chess.pgn.read_game(f)) is not None:
55
  board = game.board()
56
- base = {"1-0":1,"0-1":-1}.get(game.headers["Result"], 0)
57
  for move in game.mainline_moves():
58
  state = ChessEnv.get_piece_configuration(board)
59
  state = state if board.turn else -state
@@ -61,8 +139,8 @@ class ChessPGNDataset(GeneratorBasedBuilder):
61
  result = base * (-1 if board.turn == 0 else 1)
62
  yield uid, {
63
  "state": state.astype("int8"),
64
- "action": int(action),
65
- "result": int(result),
66
  }
67
  uid += 1
68
  board.push(move)
 
1
+ # from datasets import (
2
+ # GeneratorBasedBuilder,
3
+ # DatasetInfo,
4
+ # SplitGenerator,
5
+ # Split,
6
+ # DownloadManager,
7
+ # Features,
8
+ # Value,
9
+ # Array2D,
10
+ # )
11
+ # from adversarial_gym.chess_env import ChessEnv
12
 
13
+ # class ChessPGNDataset(GeneratorBasedBuilder):
14
+ # VERSION = "0.0.1"
15
+
16
+ # def _info(self):
17
+ # return DatasetInfo(
18
+ # description="Chess positions + moves + results, streamed from PGN shards",
19
+ # features=Features({
20
+ # "state": Array2D((8,8), dtype="int8"),
21
+ # "action": Value("int16"),
22
+ # "result": Value("int8"),
23
+ # })
24
+ # )
25
+
26
+ # def _split_generators(self, dl_manager: DownloadManager):
27
+ # from pathlib import Path
28
+
29
+ # data_dir = Path(self.config.data_dir)
30
+
31
+ # train_shards = sorted(data_dir.joinpath("train").glob("*.pgn"))
32
+ # test_shards = sorted(data_dir.joinpath("test").glob("*.pgn"))
33
+ # train_paths = dl_manager.download([str(p) for p in train_shards])
34
+ # test_paths = dl_manager.download([str(p) for p in test_shards])
35
+
36
+ # return [
37
+ # SplitGenerator(
38
+ # name=Split.TRAIN,
39
+ # gen_kwargs={"shards": train_paths},
40
+ # ),
41
+ # SplitGenerator(
42
+ # name=Split.TEST,
43
+ # gen_kwargs={"shards": test_paths},
44
+ # ),
45
+ # ]
46
+ # def _generate_examples(self, shards):
47
+ # import chess.pgn
48
+ # uid = 0
49
+
50
+ # for path in sorted(shards):
51
+ # with open(path, "r") as f:
52
+ # while (game := chess.pgn.read_game(f)) is not None:
53
+ # board = game.board()
54
+ # base = {"1-0":1,"0-1":-1}.get(game.headers["Result"], 0)
55
+ # for move in game.mainline_moves():
56
+ # state = ChessEnv.get_piece_configuration(board)
57
+ # state = state if board.turn else -state
58
+ # action = ChessEnv.move_to_action(move)
59
+ # result = base * (-1 if board.turn == 0 else 1)
60
+ # yield uid, {
61
+ # "state": state.astype("int8"),
62
+ # "action": int(action),
63
+ # "result": int(result),
64
+ # }
65
+ # uid += 1
66
+ # board.push(move)
67
+
68
+
69
+
70
+
71
+ from datasets import GeneratorBasedBuilder, DatasetInfo, SplitGenerator, Split, Features, Value, Array2D, DownloadManager
72
+
73
+
74
+ from pathlib import Path
75
+ from datasets import BuilderConfig, Version
76
+
77
+ class ChessPGNConfig(BuilderConfig):
78
+ def __init__( self,
79
+ name: str = "default",
80
+ version: Version = Version("1.0.0"),
81
+ description: str = "PGN chess shards",
82
+ data_dir: str = None,
83
+ data_files: dict = None,
84
+ **kwargs):
85
+ super().__init__(
86
+ name=name,
87
+ version=version,
88
+ description=description,
89
+ data_dir=data_dir,
90
+ data_files=data_files or {},
91
+ **kwargs
92
+ )
93
  class ChessPGNDataset(GeneratorBasedBuilder):
94
+ BUILDER_CONFIG_CLASS = ChessPGNConfig
95
+ BUILDER_CONFIGS = [
96
+ ChessPGNConfig(
97
+ name="default",
98
+ data_files={
99
+ "train": [str(p) for p in Path("train").glob("*.pgn")],
100
+ "test": [str(p) for p in Path("test").glob("*.pgn")],
101
+ },
102
+ )
103
+ ]
104
+
105
+ DEFAULT_WRITER_BATCH_SIZE = 512
106
 
107
  def _info(self):
108
  return DatasetInfo(
 
109
  features=Features({
110
  "state": Array2D((8,8), dtype="int8"),
111
  "action": Value("int16"),
112
  "result": Value("int8"),
113
+ }),
114
  )
115
 
116
  def _split_generators(self, dl_manager: DownloadManager):
117
+ # use config.data_files directly—no data_dir needed
118
+ train_files = dl_manager.download(self.config.data_files["train"])
119
+ test_files = dl_manager.download(self.config.data_files["test"])
 
 
 
 
 
 
 
 
120
  return [
121
+ SplitGenerator(name=Split.TRAIN, gen_kwargs={"shards": train_files}),
122
+ SplitGenerator(name=Split.TEST, gen_kwargs={"shards": test_files}),
 
 
 
 
 
 
123
  ]
124
+
125
  def _generate_examples(self, shards):
126
+ import chess.pgn, numpy as np
127
+ from adversarial_gym.chess_env import ChessEnv
128
 
129
+ uid = 0
130
  for path in sorted(shards):
131
+ with open(path) as f:
132
  while (game := chess.pgn.read_game(f)) is not None:
133
  board = game.board()
134
+ base = {"1-0":1,"0-1":-1}.get(game.headers["Result"],0)
135
  for move in game.mainline_moves():
136
  state = ChessEnv.get_piece_configuration(board)
137
  state = state if board.turn else -state
 
139
  result = base * (-1 if board.turn == 0 else 1)
140
  yield uid, {
141
  "state": state.astype("int8"),
142
+ "action": np.int16(action),
143
+ "result": np.int8(result),
144
  }
145
  uid += 1
146
  board.push(move)