Datasets:
Update README.md
Browse files
README.md
CHANGED
@@ -1,22 +1,73 @@
|
|
1 |
-
---
|
2 |
-
license: apache-2.0
|
3 |
-
dataset_info:
|
4 |
-
features:
|
5 |
-
- name: image
|
6 |
-
dtype: image
|
7 |
-
- name: target
|
8 |
-
dtype: string
|
9 |
-
- name: turn
|
10 |
-
dtype: string
|
11 |
-
splits:
|
12 |
-
- name: train
|
13 |
-
num_bytes: 82595245
|
14 |
-
num_examples: 3500
|
15 |
-
download_size: 81770022
|
16 |
-
dataset_size: 82595245
|
17 |
-
configs:
|
18 |
-
- config_name: default
|
19 |
-
data_files:
|
20 |
-
- split: train
|
21 |
-
path: data/train-*
|
22 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
dataset_info:
|
4 |
+
features:
|
5 |
+
- name: image
|
6 |
+
dtype: image
|
7 |
+
- name: target
|
8 |
+
dtype: string
|
9 |
+
- name: turn
|
10 |
+
dtype: string
|
11 |
+
splits:
|
12 |
+
- name: train
|
13 |
+
num_bytes: 82595245
|
14 |
+
num_examples: 3500
|
15 |
+
download_size: 81770022
|
16 |
+
dataset_size: 82595245
|
17 |
+
configs:
|
18 |
+
- config_name: default
|
19 |
+
data_files:
|
20 |
+
- split: train
|
21 |
+
path: data/train-*
|
22 |
+
task_categories:
|
23 |
+
- image-to-text
|
24 |
+
pretty_name: BIG-Bench Checkmate In One Move (Images)
|
25 |
+
---
|
26 |
+
|
27 |
+
# Dataset Card for BIG-Bench Checkmate In One Move (Images)
|
28 |
+
## Description
|
29 |
+
This is an adapted version of the [BIG-Bench Checkmate in One Move task](https://github.com/google/BIG-bench/tree/main/bigbench/benchmark_tasks/checkmate_in_one)
|
30 |
+
as originally made by Nitish Keskar (nkeskar@salesforce.com).
|
31 |
+
Copying the original task description:
|
32 |
+
>The goal of this task is to probe the ability of language models to play chess in standard algebraic notation (SAN). The input to the model is a sequence of moves such that a next possible move is a checkmate. We curate 3,500 games and measure the performance of the system in exact match accuracy.
|
33 |
+
|
34 |
+
This version simply replaces the text input with images of the board state before the checkmate move.
|
35 |
+
## Example
|
36 |
+

|
37 |
+
|
38 |
+
turn: black
|
39 |
+
|
40 |
+
target: Rg5#
|
41 |
+
|
42 |
+
## Generation Code
|
43 |
+
The code used to generate this was:
|
44 |
+
```python
|
45 |
+
def svg_to_png(svg_data, output_file):
|
46 |
+
"""Convert SVG data to PNG and save it"""
|
47 |
+
# Convert SVG to PNG using cairosvg
|
48 |
+
png_data = cairosvg.svg2png(bytestring=svg_data)
|
49 |
+
|
50 |
+
# Write the PNG data to the output file
|
51 |
+
with open(output_file, "wb") as f:
|
52 |
+
f.write(png_data)
|
53 |
+
|
54 |
+
|
55 |
+
def pgn_to_images(pgn, f_name):
|
56 |
+
"""Convert a PGN file to an image of the final board position"""
|
57 |
+
pgn = io.StringIO(pgn)
|
58 |
+
game = chess.pgn.read_game(pgn)
|
59 |
+
board = game.board()
|
60 |
+
for move in game.mainline_moves():
|
61 |
+
board.push(move)
|
62 |
+
svg = chess.svg.board(board)
|
63 |
+
turn_str = "white" if board.turn == chess.WHITE else "black"
|
64 |
+
# use a temporary file to convert the SVG to PNG
|
65 |
+
with open("temp.svg", "w") as f:
|
66 |
+
f.write(svg)
|
67 |
+
with open("temp.svg", "rb") as f:
|
68 |
+
svg_data = f.read()
|
69 |
+
svg_to_png(svg_data, f_name)
|
70 |
+
os.remove("temp.svg")
|
71 |
+
return turn_str
|
72 |
+
```
|
73 |
+
Thanks to the python chess library and cairosvg for making this possible.
|