Upload prepare.py
Browse files- prepare.py +145 -0
prepare.py
ADDED
|
@@ -0,0 +1,145 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os
|
| 2 |
+
import re
|
| 3 |
+
import io
|
| 4 |
+
import requests
|
| 5 |
+
from zipfile import ZipFile
|
| 6 |
+
from tqdm import tqdm
|
| 7 |
+
import chess.pgn as pgn
|
| 8 |
+
import pandas as pd
|
| 9 |
+
from datasets import Dataset, DatasetInfo
|
| 10 |
+
|
| 11 |
+
tqdm.pandas()
|
| 12 |
+
|
| 13 |
+
ZIP_URLS = [
|
| 14 |
+
"https://database.nikonoel.fr/lichess_elite_2023-01.zip",
|
| 15 |
+
"https://database.nikonoel.fr/lichess_elite_2023-02.zip",
|
| 16 |
+
"https://database.nikonoel.fr/lichess_elite_2023-03.zip",
|
| 17 |
+
"https://database.nikonoel.fr/lichess_elite_2023-04.zip",
|
| 18 |
+
"https://database.nikonoel.fr/lichess_elite_2023-05.zip",
|
| 19 |
+
"https://database.nikonoel.fr/lichess_elite_2023-06.zip",
|
| 20 |
+
"https://database.nikonoel.fr/lichess_elite_2023-07.zip",
|
| 21 |
+
"https://database.nikonoel.fr/lichess_elite_2023-08.zip",
|
| 22 |
+
"https://database.nikonoel.fr/lichess_elite_2023-09.zip",
|
| 23 |
+
"https://database.nikonoel.fr/lichess_elite_2023-10.zip",
|
| 24 |
+
"https://database.nikonoel.fr/lichess_elite_2023-11.zip",
|
| 25 |
+
"https://database.nikonoel.fr/lichess_elite_2023-12.zip",
|
| 26 |
+
]
|
| 27 |
+
|
| 28 |
+
|
| 29 |
+
def download_and_unzip(url, save_directory, force_download=False):
|
| 30 |
+
# Extract the filename from the URL
|
| 31 |
+
filename = url.split("/")[-1]
|
| 32 |
+
file_path = os.path.join(save_directory, filename)
|
| 33 |
+
|
| 34 |
+
# Check if the file already exists
|
| 35 |
+
if force_download or not os.path.exists(file_path):
|
| 36 |
+
# Download the file if it does not exist
|
| 37 |
+
print(f"Downloading {filename}...")
|
| 38 |
+
response = requests.get(url)
|
| 39 |
+
with open(file_path, "wb") as file:
|
| 40 |
+
file.write(response.content)
|
| 41 |
+
print(f"Downloaded {filename}.")
|
| 42 |
+
else:
|
| 43 |
+
print(f"{filename} already exists. Skipping download.")
|
| 44 |
+
|
| 45 |
+
# Unzip the file
|
| 46 |
+
with ZipFile(file_path, "r") as zip_ref:
|
| 47 |
+
print(f"Unzipping {filename}...")
|
| 48 |
+
zip_ref.extractall(save_directory)
|
| 49 |
+
print(f"Unzipped {filename}.")
|
| 50 |
+
|
| 51 |
+
|
| 52 |
+
def parse_pgn_dataset_to_dataframe(pgn_file_path):
|
| 53 |
+
# Regular expressions for matching headers and moves
|
| 54 |
+
header_pattern = re.compile(r"\[([A-Za-z0-9]+) \"(.+?)\"\]")
|
| 55 |
+
|
| 56 |
+
games_list = []
|
| 57 |
+
current_game = {}
|
| 58 |
+
transcript = []
|
| 59 |
+
|
| 60 |
+
with open(pgn_file_path, "r") as file:
|
| 61 |
+
for line in file:
|
| 62 |
+
line = line.encode("utf-8").decode("ascii", "ignore")
|
| 63 |
+
header_match = header_pattern.match(line)
|
| 64 |
+
if header_match:
|
| 65 |
+
# If a new game starts and current_game is not empty, save the current game
|
| 66 |
+
if header_match.group(1) == "Event" and current_game:
|
| 67 |
+
current_game["transcript"] = " ".join(transcript).strip()
|
| 68 |
+
games_list.append(current_game)
|
| 69 |
+
current_game = {}
|
| 70 |
+
transcript = []
|
| 71 |
+
current_game[header_match.group(1)] = header_match.group(2)
|
| 72 |
+
else:
|
| 73 |
+
# Add moves to transcript, ignoring empty lines and game results
|
| 74 |
+
clean_line = line.strip()
|
| 75 |
+
if (
|
| 76 |
+
clean_line
|
| 77 |
+
and not clean_line.startswith("1-0")
|
| 78 |
+
and not clean_line.startswith("1/2-1/2")
|
| 79 |
+
and not clean_line.startswith("0-1")
|
| 80 |
+
):
|
| 81 |
+
transcript.append(clean_line)
|
| 82 |
+
|
| 83 |
+
# Add the last game if it exists
|
| 84 |
+
if current_game:
|
| 85 |
+
current_game["transcript"] = " ".join(transcript).strip()
|
| 86 |
+
games_list.append(current_game)
|
| 87 |
+
|
| 88 |
+
# Create a DataFrame
|
| 89 |
+
df = pd.DataFrame(games_list)
|
| 90 |
+
return df
|
| 91 |
+
|
| 92 |
+
|
| 93 |
+
def pgn_to_uci_transcript(pgn_transcript):
|
| 94 |
+
game = pgn.read_game(io.StringIO(pgn_transcript))
|
| 95 |
+
if game is None:
|
| 96 |
+
return
|
| 97 |
+
|
| 98 |
+
board = game.board()
|
| 99 |
+
move_list = []
|
| 100 |
+
for move in game.mainline_moves():
|
| 101 |
+
move_list.append(board.uci(move))
|
| 102 |
+
board.push(move)
|
| 103 |
+
|
| 104 |
+
return " ".join(move_list)
|
| 105 |
+
|
| 106 |
+
|
| 107 |
+
if __name__ == "__main__":
|
| 108 |
+
save_directory = "."
|
| 109 |
+
|
| 110 |
+
if not os.path.exists(save_directory):
|
| 111 |
+
os.makedirs(save_directory)
|
| 112 |
+
|
| 113 |
+
for url in ZIP_URLS:
|
| 114 |
+
download_and_unzip(url, save_directory)
|
| 115 |
+
|
| 116 |
+
pgn_files = [file for file in os.listdir(
|
| 117 |
+
save_directory) if file.endswith(".pgn")]
|
| 118 |
+
|
| 119 |
+
file_dfs = []
|
| 120 |
+
|
| 121 |
+
for pgn_file in pgn_files:
|
| 122 |
+
print(f"Parsing PGN from: {pgn_file}")
|
| 123 |
+
df = parse_pgn_dataset_to_dataframe(pgn_file)
|
| 124 |
+
df = df[df["EventType"] == "rapid"]
|
| 125 |
+
|
| 126 |
+
file_dfs.append(df)
|
| 127 |
+
|
| 128 |
+
df = pd.concat(file_dfs)
|
| 129 |
+
|
| 130 |
+
# cast numeric columns
|
| 131 |
+
for column in df.columns:
|
| 132 |
+
if df[column].str.isnumeric().all():
|
| 133 |
+
df[column] = df[column].astype(int)
|
| 134 |
+
|
| 135 |
+
df["transcript"] = df["transcript"].progress_apply(pgn_to_uci_transcript)
|
| 136 |
+
|
| 137 |
+
# filter unresolved games
|
| 138 |
+
df = df[df["Result"] != "*"]
|
| 139 |
+
|
| 140 |
+
df.to_feather("elite_dataset.feather")
|
| 141 |
+
ds_info = DatasetInfo(
|
| 142 |
+
description="The Lichess Elite Dataset includes all (rapid) games from Lichess by players rated 2500+ against players rated 2300+ played during the year 2023. Only games with an outcome of 1/2-1/2, 1-0, or 0-1 are included."
|
| 143 |
+
)
|
| 144 |
+
ds = Dataset.from_pandas(df, info=ds_info)
|
| 145 |
+
ds.push_to_hub("austindavis/chess_world_lichess_elite")
|