Upload lichess_sf_hf.py
Browse files- lichess_sf_hf.py +106 -0
lichess_sf_hf.py
ADDED
@@ -0,0 +1,106 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#%%
|
2 |
+
import datasets
|
3 |
+
import zstandard as zstd
|
4 |
+
import io
|
5 |
+
|
6 |
+
#%%
|
7 |
+
class LichessConfig(datasets.BuilderConfig):
|
8 |
+
def __init__(self, features, **kwargs):
|
9 |
+
super(LichessConfig, self).__init__(**kwargs)
|
10 |
+
self.features = features
|
11 |
+
|
12 |
+
|
13 |
+
class Lichess(datasets.GeneratorBasedBuilder):
|
14 |
+
BUILDER_CONFIG_CLASS = LichessConfig
|
15 |
+
BUILDER_CONFIGS = [LichessConfig(name="pgn",
|
16 |
+
features=["WhiteElo",
|
17 |
+
"BlackElo",
|
18 |
+
"pgn"]),
|
19 |
+
|
20 |
+
LichessConfig(name="fen",
|
21 |
+
features=["WhiteElo",
|
22 |
+
"BlackElo",
|
23 |
+
"fens",
|
24 |
+
"moves",
|
25 |
+
"scores"]),]
|
26 |
+
|
27 |
+
def _info(self):
|
28 |
+
features_dict = {feature: datasets.Value("uint16") for feature in self.config.features}
|
29 |
+
|
30 |
+
if self.config.name == "pgn":
|
31 |
+
features_dict["pgn"] = datasets.Value("string")
|
32 |
+
|
33 |
+
|
34 |
+
if self.config.name == "fen":
|
35 |
+
features_dict["fens"] = datasets.Value("string")
|
36 |
+
features_dict["moves"] = datasets.Value("string")
|
37 |
+
features_dict["scores"] = datasets.Value("string")
|
38 |
+
|
39 |
+
info = datasets.DatasetInfo(datasets.Features(features_dict))
|
40 |
+
return info
|
41 |
+
|
42 |
+
def _get_filepaths(self):
|
43 |
+
#months = ["01", "02", "03", "04", "05", "06",
|
44 |
+
# "07", "08", "09", "10", "11", "12"]
|
45 |
+
months = ["04"]
|
46 |
+
shards = ["0", "1", "2", "3"]
|
47 |
+
filepaths = ["lichess/" + self.config.name + "/2023/" + m for m in months]
|
48 |
+
|
49 |
+
if self.config.name == "pgn":
|
50 |
+
paths = []
|
51 |
+
for shard in shards:
|
52 |
+
paths.extend([filepath + "/" + shard + ".zst" for filepath in filepaths])
|
53 |
+
|
54 |
+
if self.config.name == "fen":
|
55 |
+
paths = []
|
56 |
+
for shard in shards:
|
57 |
+
paths.extend([filepath + "/" + shard + "_fen.zst" for filepath in filepaths])
|
58 |
+
|
59 |
+
return paths
|
60 |
+
|
61 |
+
def _split_generators(self, dl_manager: datasets.DownloadManager):
|
62 |
+
filepaths = self._get_filepaths()
|
63 |
+
downloaded_files = dl_manager.download(filepaths)
|
64 |
+
generator = datasets.SplitGenerator(name=datasets.Split.TRAIN,
|
65 |
+
gen_kwargs={'filepaths': downloaded_files})
|
66 |
+
return [generator]
|
67 |
+
|
68 |
+
def _generate_examples(self, filepaths):
|
69 |
+
dctx = zstd.ZstdDecompressor()
|
70 |
+
for filepath in filepaths:
|
71 |
+
n = 0
|
72 |
+
with open(filepath, "rb") as file:
|
73 |
+
with dctx.stream_reader(file) as sr:
|
74 |
+
pgn = io.TextIOWrapper(sr)
|
75 |
+
|
76 |
+
if self.config.name == "pgn":
|
77 |
+
while True:
|
78 |
+
white_elo, black_elo = pgn.readline().split(" ")
|
79 |
+
game_pgn = pgn.readline()
|
80 |
+
next_line = pgn.readline()
|
81 |
+
|
82 |
+
if game_pgn:
|
83 |
+
_id = n
|
84 |
+
n += 1
|
85 |
+
yield _id, {"WhiteElo": int(white_elo),
|
86 |
+
"BlackElo": int(black_elo),
|
87 |
+
"pgn": game_pgn}
|
88 |
+
else:
|
89 |
+
break
|
90 |
+
|
91 |
+
elif self.config.name == "fen":
|
92 |
+
while True:
|
93 |
+
white_elo, black_elo = pgn.readline().split(" ")
|
94 |
+
game = pgn.readline()
|
95 |
+
fens, moves, scores = game.split(";")
|
96 |
+
|
97 |
+
if game:
|
98 |
+
_id = n
|
99 |
+
n += 1
|
100 |
+
yield _id, {"WhiteElo": white_elo,
|
101 |
+
"BlackElo": black_elo,
|
102 |
+
"fens": fens,
|
103 |
+
"moves": moves,
|
104 |
+
"scores": scores}
|
105 |
+
else:
|
106 |
+
break
|