mirror of
https://github.com/open-thought/reasoning-gym.git
synced 2026-04-19 12:58:07 +00:00
add bitwise arithmetic
This commit is contained in:
parent
bedee59616
commit
17088e9b42
2 changed files with 207 additions and 0 deletions
54
tests/test_bitwise_arithmetic.py
Normal file
54
tests/test_bitwise_arithmetic.py
Normal file
|
|
@ -0,0 +1,54 @@
|
|||
import pytest
|
||||
|
||||
from reasoning_gym.arithmetic.bitwise_arithmetic import BitwiseArithmeticConfig, BitwiseArithmeticDataset
|
||||
|
||||
|
||||
def test_bitwise_arithmetic():
|
||||
"""Test basic properties and solution of generated items"""
|
||||
|
||||
# Easy
|
||||
config = BitwiseArithmeticConfig(
|
||||
seed=42,
|
||||
size=2000,
|
||||
difficulty=1,
|
||||
)
|
||||
dataset = BitwiseArithmeticDataset(config)
|
||||
|
||||
for item in dataset:
|
||||
assert isinstance(item, dict)
|
||||
assert "question" in item
|
||||
assert "answer" in item
|
||||
assert "metadata" in item
|
||||
|
||||
# Test the scoring
|
||||
assert dataset.score_answer(answer=item["answer"], entry=item) == 1.0
|
||||
assert dataset.score_answer(answer=None, entry=item) == 0.0
|
||||
assert dataset.score_answer(answer="kitty cat", entry=item) == 0.01
|
||||
|
||||
config = BitwiseArithmeticConfig(
|
||||
seed=42,
|
||||
size=100,
|
||||
difficulty=5,
|
||||
)
|
||||
dataset = BitwiseArithmeticDataset(config)
|
||||
|
||||
for item in dataset:
|
||||
assert isinstance(item, dict)
|
||||
assert "question" in item
|
||||
assert "answer" in item
|
||||
assert "metadata" in item
|
||||
assert dataset.score_answer(answer=item["answer"], entry=item) == 1.0
|
||||
|
||||
config = BitwiseArithmeticConfig(
|
||||
seed=42,
|
||||
size=100,
|
||||
difficulty=10,
|
||||
)
|
||||
dataset = BitwiseArithmeticDataset(config)
|
||||
|
||||
for item in dataset:
|
||||
assert isinstance(item, dict)
|
||||
assert "question" in item
|
||||
assert "answer" in item
|
||||
assert "metadata" in item
|
||||
assert dataset.score_answer(answer=item["answer"], entry=item) == 1.0
|
||||
Loading…
Add table
Add a link
Reference in a new issue