mirror of
https://github.com/open-thought/reasoning-gym.git
synced 2026-04-19 12:58:07 +00:00
Add Modulo Grid Task (#273)
* add modulo_grid dataset * ensure the pattern is mathematical, not just spatial --------- Co-authored-by: Andreas Koepf <andreas.koepf@provisio.com>
This commit is contained in:
parent
07dc01ad87
commit
cbfdf097a0
3 changed files with 182 additions and 0 deletions
37
tests/test_modulo_grid.py
Normal file
37
tests/test_modulo_grid.py
Normal file
|
|
@ -0,0 +1,37 @@
|
|||
import pytest
|
||||
|
||||
from reasoning_gym.cognition.modulo_grid import ModuloGridConfig, ModuloGridDataset
|
||||
|
||||
|
||||
def test_modulo_grid():
|
||||
"""Test basic properties and solution of generated items"""
|
||||
|
||||
# Easy
|
||||
config = ModuloGridConfig(seed=42, size=50, size_x=10, size_y=10, max_divisor=10, max_target=10, max_holes=1)
|
||||
dataset = ModuloGridDataset(config)
|
||||
|
||||
for item in dataset:
|
||||
assert isinstance(item, dict)
|
||||
assert "question" in item and isinstance(item["question"], str)
|
||||
assert "answer" in item and isinstance(item["answer"], str)
|
||||
assert "metadata" in item
|
||||
|
||||
# Test the scoring
|
||||
assert item["question"] != item["answer"]
|
||||
assert dataset.score_answer(answer=item["answer"], entry=item) == 1.0
|
||||
assert dataset.score_answer(answer=None, entry=item) == 0.0
|
||||
|
||||
# Hard
|
||||
config = ModuloGridConfig(seed=42, size=50, size_x=25, size_y=25, max_divisor=25, max_target=25, max_holes=15)
|
||||
dataset = ModuloGridDataset(config)
|
||||
|
||||
for item in dataset:
|
||||
assert isinstance(item, dict)
|
||||
assert "question" in item and isinstance(item["question"], str)
|
||||
assert "answer" in item and isinstance(item["answer"], str)
|
||||
assert "metadata" in item
|
||||
|
||||
# Test the scoring
|
||||
assert item["question"] != item["answer"]
|
||||
assert dataset.score_answer(answer=item["answer"], entry=item) == 1.0
|
||||
assert dataset.score_answer(answer=None, entry=item) == 0.0
|
||||
Loading…
Add table
Add a link
Reference in a new issue