mirror of
https://github.com/open-thought/reasoning-gym.git
synced 2026-04-25 17:10:51 +00:00
80 lines
3.4 KiB
Python
80 lines
3.4 KiB
Python
from random import Random
|
|
from typing import Dict, Any
|
|
|
|
def generate_from_variables(name: str, item1: str, item2: str, price1: int, price2: int,
|
|
total: float, n1: int, percent: int, currency: str) -> Dict[str, Any]:
|
|
change = total * percent / 100
|
|
spent = total * (100 - percent) / 100
|
|
cost_item1 = n1 * price1 / 100
|
|
spent_item2 = spent - cost_item1
|
|
n2 = int(spent_item2 / (price2 / 100))
|
|
|
|
question = f"The vending machines sell {item1} for {price1} cents and {item2} for {price2} cents. {name} spent {currency}{total} and got {n1} bags of {item1} and had {percent}% of his money left. How many {item2} did he buy?"
|
|
|
|
answer_cot = f"{name} got {currency}{change} in change because {total} x {percent}/100 = {change}\n" \
|
|
f"{name} spent {currency}{spent} because {total} - {change} = {spent}\n" \
|
|
f"{name} spent {currency}{cost_item1} on {item1} because {n1} x {price1/100} = {cost_item1}\n" \
|
|
f"{name} spent {spent_item2} on {item2} because {spent} - {cost_item1} = {spent_item2}\n" \
|
|
f"{name} bought {n2} {item2} because {spent_item2} / {price2/100} = {n2}\n" \
|
|
f"#### {n2}"
|
|
|
|
return {
|
|
'question': question,
|
|
'answer': str(n2),
|
|
'answer_cot': answer_cot,
|
|
'answer_value': n2,
|
|
'variables': {
|
|
'name': name,
|
|
'item1': item1,
|
|
'item2': item2,
|
|
'price1': price1,
|
|
'price2': price2,
|
|
'total_spent': total,
|
|
'num_item1': n1,
|
|
'num_item2': n2,
|
|
'percent_change': percent,
|
|
'currency': currency
|
|
}
|
|
}
|
|
|
|
def generate_example(rng: Random, difficulty: float = 1.0) -> Dict[str, Any]:
|
|
names = ["George", "James", "John", "Robert", "Michael", "William"]
|
|
items = ['pretzels', 'popcorn', 'gum', 'cookies', 'crackers']
|
|
currencies = ['$', '£', '€']
|
|
|
|
name = rng.choice(names)
|
|
item1, item2 = rng.sample(items, 2)
|
|
currency = rng.choice(currencies)
|
|
|
|
price1 = int(rng.randrange(25, int(100 * difficulty), 5))
|
|
price2 = int(rng.randrange(50, int(150 * difficulty), 5))
|
|
while price2 <= price1:
|
|
price2 = int(rng.randrange(50, int(150 * difficulty), 5))
|
|
|
|
total = int(rng.randrange(500, int(1500 * difficulty), 100))
|
|
n1 = int(rng.randint(1, int(10 * difficulty)))
|
|
percent = int(rng.randint(1, int(10 * difficulty)))
|
|
|
|
# Validate conditions
|
|
while not (isinstance(total * percent / 100, int) and
|
|
isinstance((total * (100 - percent) / 100 - n1 * price1 / 100) / (price2 / 100), int) and
|
|
(total * (100 - percent) / 100 - n1 * price1 / 100) % (price2 / 100) == 0):
|
|
total = int(rng.randrange(500, int(1500 * difficulty), 100))
|
|
n1 = int(rng.randint(1, int(10 * difficulty)))
|
|
percent = int(rng.randint(1, int(10 * difficulty)))
|
|
|
|
result = generate_from_variables(name, item1, item2, price1, price2, total, n1, percent, currency)
|
|
|
|
return {
|
|
'question': result['question'],
|
|
'answer': result['answer'],
|
|
'metadata': {
|
|
'difficulty': difficulty,
|
|
'answer_value': result['answer_value'],
|
|
'answer_cot': result['answer_cot'],
|
|
'variables': result['variables']
|
|
}
|
|
}
|
|
|
|
def original_example() -> Dict[str, Any]:
|
|
return generate_from_variables("George", "chips", "candy bars", 40, 75, 5, 3, 1, "$")
|