mirror of
https://github.com/open-thought/reasoning-gym.git
synced 2026-04-25 17:10:51 +00:00
83 lines
3.7 KiB
Python
83 lines
3.7 KiB
Python
from random import Random
|
|
from typing import Dict, Any
|
|
|
|
def generate_from_variables(name: str, num_bills: int, bill_value: int, num_items1: int,
|
|
price1: int, num_items2: int, price2: int, item1: str,
|
|
item2: str, currency: str) -> Dict[str, Any]:
|
|
|
|
initial_amount = num_bills * bill_value
|
|
spent_items1 = num_items1 * price1
|
|
spent_items2 = num_items2 * price2
|
|
total_spent = spent_items1 + spent_items2
|
|
remaining = initial_amount - total_spent
|
|
|
|
question = f"{name} has {num_bills} {currency}{bill_value} bills. He buys {num_items1} {item1}s for {currency}{price1} each. He also buys {num_items2} packs of {item2}s for {currency}{price2} each. How much money does he have left?"
|
|
|
|
answer_cot = f"{name} starts off with {num_bills} * {currency}{bill_value} = {currency}{initial_amount}.\n" \
|
|
f"{name} spends {num_items1} {item1}s * {currency}{price1} = {currency}{spent_items1} on {item1}s.\n" \
|
|
f"{name} spends {num_items2} packs of {item2}s * {currency}{price2} = {currency}{spent_items2} on {item2}s.\n" \
|
|
f"Total {name} has spent {currency}{spent_items1} + {currency}{spent_items2} = {currency}{total_spent}.\n" \
|
|
f"{name} has {currency}{initial_amount} - {currency}{total_spent} = {currency}{remaining} remaining.\n#### {remaining}"
|
|
|
|
return {
|
|
'question': question,
|
|
'answer': str(remaining),
|
|
'answer_cot': answer_cot,
|
|
'answer_value': remaining,
|
|
'variables': {
|
|
'name': name,
|
|
'num_bills': num_bills,
|
|
'bill_value': bill_value,
|
|
'num_items1': num_items1,
|
|
'price1': price1,
|
|
'num_items2': num_items2,
|
|
'price2': price2,
|
|
'item1': item1,
|
|
'item2': item2,
|
|
'currency': currency,
|
|
'initial_amount': initial_amount,
|
|
'total_spent': total_spent
|
|
}
|
|
}
|
|
|
|
def generate_example(rng: Random, difficulty: float = 1.0) -> Dict[str, Any]:
|
|
names = ["Craig", "John", "Michael", "David", "James", "Robert", "William"]
|
|
items1 = ["toy car", "action figure", "coloring book", "puzzle", "board game"]
|
|
items2 = ["sticker", "candy bar", "trading card", "pencil", "eraser"]
|
|
currencies = ["$", "€", "£"]
|
|
bills = [(5, 5), (10, 10), (20, 20), (50, 50), (100, 100)]
|
|
|
|
name = rng.choice(names)
|
|
item1 = rng.choice(items1)
|
|
item2 = rng.choice(items2)
|
|
currency = rng.choice(currencies)
|
|
bill_value = rng.choice(bills)[1]
|
|
|
|
num_bills = int(rng.randint(1, int(10 * difficulty)))
|
|
num_items1 = int(rng.randint(2, int(15 * difficulty)))
|
|
num_items2 = int(rng.randint(2, int(10 * difficulty)))
|
|
price1 = int(rng.randint(1, int(10 * difficulty)))
|
|
price2 = int(rng.randint(1, int(10 * difficulty)))
|
|
|
|
# Ensure total cost doesn't exceed available money
|
|
while (num_items1 * price1 + num_items2 * price2) > (num_bills * bill_value):
|
|
num_items1 = int(rng.randint(2, int(15 * difficulty)))
|
|
num_items2 = int(rng.randint(2, int(10 * difficulty)))
|
|
|
|
result = generate_from_variables(name, num_bills, bill_value, num_items1, price1,
|
|
num_items2, price2, item1, item2, currency)
|
|
|
|
return {
|
|
'question': result['question'],
|
|
'answer': result['answer'],
|
|
'metadata': {
|
|
'difficulty': difficulty,
|
|
'answer_value': result['answer_value'],
|
|
'answer_cot': result['answer_cot'],
|
|
'variables': result['variables']
|
|
}
|
|
}
|
|
|
|
def original_example() -> Dict[str, Any]:
|
|
return generate_from_variables("Craig", 2, 20, 6, 2, 3, 3,
|
|
"squirt gun", "water balloon", "$")
|