Create python generator files for gsm symbolic templates

reasoning_gym/arithmetic/gsm_symbolic/generator_23.py

@@ -0,0 +1,70 @@
+from random import Random
+from typing import Dict, Any
+
+def generate_from_variables(name: str, task: str, profession: str, hours: int, 
+                          work_type: str, rate: int, fee: int, currency: str) -> Dict[str, Any]:
+    lost_income = hours * rate
+    savings = lost_income - fee
+    
+    question = f"{name} is trying to decide whether to do {task} herself or hire an {profession}. If she does it herself, she'll be able to do {hours} fewer hours of {work_type} work, losing {currency}{rate}/hour in missed income. The {profession} charges {currency}{fee}. How much more money will she have if she hires the {profession}?"
+    
+    answer_cot = f"First find the total lost revenue if {name} does {task} herself: {currency}{rate}/hour * {hours} hours = {currency}{lost_income}\nThen subtract the {profession}'s charge to find how much money {name} saves: {currency}{lost_income} - {currency}{fee} = {currency}{savings}\n#### {savings}"
+
+    return {
+        'question': question,
+        'answer': str(savings),
+        'answer_cot': answer_cot,
+        'answer_value': savings,
+        'variables': {
+            'name': name,
+            'task': task,
+            'profession': profession,
+            'hours': hours,
+            'work_type': work_type,
+            'hourly_rate': rate,
+            'fee': fee,
+            'currency': currency,
+            'lost_income': lost_income
+        }
+    }
+
+def generate_example(rng: Random, difficulty: float = 1.0) -> Dict[str, Any]:
+    names_female = ['Emma', 'Sophia', 'Isabella', 'Olivia', 'Ava', 'Mia', 'Emily']
+    tasks = ["her taxes", "her financial planning", "her business accounting"]
+    professions = ["accountant", "financial advisor", "tax consultant", "bookkeeper"]
+    work_types = ["freelance", "consulting", "part-time", "contract"]
+    currencies = ['$', '€', '£']
+
+    name = rng.choice(names_female)
+    task = rng.choice(tasks)
+    profession = rng.choice(professions)
+    work_type = rng.choice(work_types)
+    currency = rng.choice(currencies)
+    
+    hours = int(rng.randint(4, int(14 * difficulty)))
+    rate = int(rng.randint(20, int(100 * difficulty)))
+    fee = int(rng.randint(50, int(200 * difficulty)))
+    
+    # Ensure conditions are met
+    while hours * rate <= fee:
+        hours = int(rng.randint(4, int(14 * difficulty)))
+        rate = int(rng.randint(20, int(100 * difficulty)))
+        fee = int(rng.randint(50, int(200 * difficulty)))
+
+    result = generate_from_variables(name, task, profession, hours, work_type, 
+                                   rate, fee, currency)
+    
+    return {
+        'question': result['question'],
+        'answer': result['answer'], 
+        'metadata': {
+            'difficulty': difficulty,
+            'answer_value': result['answer_value'],
+            'answer_cot': result['answer_cot'],
+            'variables': result['variables']
+        }
+    }
+
+def original_example() -> Dict[str, Any]:
+    return generate_from_variables("Jackie", "her taxes", "accountant", 3, 
+                                 "freelance", 35, 90, "$")