mirror of
https://github.com/InternLM/InternBootcamp.git
synced 2026-04-30 17:40:42 +00:00
298 lines
11 KiB
Python
Executable file
298 lines
11 KiB
Python
Executable file
"""# 谜题训练场开发任务
|
||
|
||
## 任务概述
|
||
你是一位资深程序员，我需要你帮我实现一个特定谜题的训练场环境类。这个类继承自`Basebootcamp`，用于生成谜题实例并验证解答。
|
||
|
||
## 背景说明
|
||
我正在开发一系列谜题训练场，每个训练场对应一个特定类型的谜题。训练场类命名为`{PuzzleName}bootcamp`，其中`PuzzleName`是谜题的名称。
|
||
|
||
每个训练场类主要提供两个核心功能：
|
||
1. 生成该谜题类型的问题实例
|
||
2. 验证用户对问题的回答是否正确
|
||
|
||
## 技术接口规范
|
||
|
||
### 类方法实现要求
|
||
|
||
```python
|
||
class {PuzzleName}bootcamp(Basebootcamp):
|
||
def __init__(self, **params):
|
||
\"\"\"
|
||
请你自定义params，以保存该puzzle相关的参数，例如网格大小等，参数配有默认值
|
||
\"\"\"
|
||
pass
|
||
|
||
def case_generator(self):
|
||
\"\"\"
|
||
生成谜题实例，提示：为保证谜题有解，可以先生成结果再对结果处理得到谜题
|
||
返回：一个可JSON序列化的字典（避免包含set等无法通过json.dumps处理的数据结构）
|
||
\"\"\"
|
||
pass
|
||
|
||
@staticmethod
|
||
def prompt_func(question_case) -> str:
|
||
\"\"\"
|
||
将case_generator生成的谜题实例转换为文本形式的问题，问题中包含问题背景、对谜题规则的介绍、具体要解决的谜题实例、期望最终答案的格式，
|
||
例如：你是xxxx，请你解答yyyy，规则如下：yyyy，最终答案放置在：zzzzz
|
||
|
||
参数:
|
||
question_case: 由case_generator生成的谜题实例
|
||
|
||
返回:
|
||
str: 格式化的问题字符串
|
||
|
||
注意:
|
||
1. 需考虑问题的格式，以便后续能正确提取
|
||
2. 问题描述中应包含期望的答案格式说明，以便后续能正确提取，为了避免抽取时匹配出干扰项，请要求模型将答案放在特定标签，如[answer] [/answer]内
|
||
\"\"\"
|
||
pass
|
||
|
||
@staticmethod
|
||
def extract_output(output):
|
||
\"\"\"
|
||
从LLM的回复中提取符合格式要求的答案，如有多个，请抽取最后一个，避免使用re.search等只抽取第一个结果的方式。
|
||
|
||
参数:
|
||
output: LLM的完整输出（包含原始问题和回答）
|
||
|
||
返回:
|
||
提取的答案，若未找到符合格式的答案则返回None
|
||
\"\"\"
|
||
pass
|
||
|
||
@classmethod
|
||
def _verify_correction(cls, solution, identity):
|
||
\"\"\"
|
||
验证提取的答案是否正确，注意一个问题可以能有多个解，按照谜题规则进行检验，不要直接匹配可能的答案。
|
||
|
||
参数:
|
||
solution: extract_output提取的答案
|
||
identity: case_generator生成的谜题实例
|
||
|
||
返回:
|
||
bool: 答案是否正确
|
||
\"\"\"
|
||
pass
|
||
```
|
||
|
||
### 验证评分方法（基类已实现）
|
||
|
||
```python
|
||
@classmethod
|
||
def verify_score(cls, model_output, identity:dict, format_score=0.1) -> float:
|
||
\"\"\"
|
||
验证输出结果并评分。
|
||
|
||
参数:
|
||
model_output: 模型的完整输出
|
||
identity: 谜题实例（由case_generator生成）
|
||
format_score: 答案格式正确时的基础分数
|
||
|
||
返回:
|
||
float: 评分结果（0-1之间）
|
||
\"\"\"
|
||
score = 0.
|
||
try:
|
||
extract_solution = cls.extract_output(model_output)
|
||
if extract_solution is None:
|
||
return score
|
||
else:
|
||
score = format_score # 格式正确时的基础分数
|
||
if cls._verify_correction(extract_solution, identity):
|
||
score = 1. # 答案完全正确时的满分
|
||
except Exception as e:
|
||
# 处理异常情况
|
||
pass
|
||
return score
|
||
```
|
||
|
||
### 使用示例
|
||
|
||
```python
|
||
# 初始化谜题训练场
|
||
bootcamp = Puzzlebootcamp()
|
||
|
||
# 生成谜题实例
|
||
case = bootcamp.case_generator()
|
||
|
||
# 将谜题转换为文本问题
|
||
prompt = Puzzlebootcamp.prompt_func(case)
|
||
|
||
# 获取LLM对问题的解答
|
||
response = get_response(prompt, \"LLM\")
|
||
|
||
# 从完整对话中提取答案
|
||
extracted_output = Puzzlebootcamp.extract_output(prompt + response)
|
||
|
||
# 验证答案并评分
|
||
score = Puzzlebootcamp.verify_score(extracted_output, case)
|
||
```
|
||
|
||
## 你的任务
|
||
请根据以下谜题描述（谜题描述可能不完整，请先结合你的知识澄清规则），实现一个完整的谜题训练场类：
|
||
|
||
### 谜题描述
|
||
|
||
**Thermometers Puzzle Rules (General Explanation):**
|
||
|
||
1. **Grid Structure**: The puzzle is played on a square grid (size can vary). Each cell may belong to one or more \"thermometers\" represented by connected paths starting at a bulb (●) and ending at a tip. Thermometers may overlap or intersect.
|
||
|
||
2. **Number Placement**:
|
||
- Fill the grid with numbers such that each row and column contains all integers from 1 to *N* (where *N* is the grid size), with no repetition (like a Latin square).
|
||
- **Exception**: Some variants may omit row/column uniqueness, depending on the puzzle's design.
|
||
|
||
3. **Thermometer Constraint**:
|
||
- For every thermometer, the numbers must **strictly increase** from the bulb to the tip. Each subsequent cell along the thermometer's path must be a larger number than the preceding cell.
|
||
|
||
4. **Clues**:
|
||
- Pre-filled numbers (if provided) must remain unchanged and act as constraints for solving.
|
||
|
||
**Objective**: Fill the grid while satisfying all row/column uniqueness rules (if applicable) and ensuring all thermometer paths follow the strict increasing order.
|
||
|
||
|
||
请完成上述谜题的训练场环境类实现，包括所有必要的方法。
|
||
"""
|
||
|
||
from bootcamp import Basebootcamp
|
||
import re
|
||
import random
|
||
|
||
class Thermometersbootcamp(Basebootcamp):
|
||
def __init__(self, size=5, num_thermometers=3, enforce_latin=True, max_retries=100):
|
||
self.size = size
|
||
self.num_thermometers = num_thermometers
|
||
self.enforce_latin = enforce_latin
|
||
self.max_retries = max_retries
|
||
|
||
def case_generator(self):
|
||
if self.enforce_latin:
|
||
solution = self._generate_latin_square(self.size)
|
||
else:
|
||
raise NotImplementedError("Non-Latin square puzzles are not supported yet.")
|
||
|
||
thermometers = []
|
||
for _ in range(self.num_thermometers):
|
||
thermometer = None
|
||
for _ in range(self.max_retries):
|
||
start_row = random.randint(0, self.size-1)
|
||
start_col = random.randint(0, self.size-1)
|
||
path = self._generate_thermometer_path(solution, start_row, start_col)
|
||
if path:
|
||
thermometer = {'bulb': path[0], 'path': path}
|
||
break
|
||
if not thermometer:
|
||
raise ValueError(f"Failed to generate thermometer after {self.max_retries} attempts")
|
||
thermometers.append(thermometer)
|
||
|
||
return {
|
||
"size": self.size,
|
||
"thermometers": thermometers,
|
||
"enforce_latin": self.enforce_latin
|
||
}
|
||
|
||
@staticmethod
|
||
def _generate_latin_square(size):
|
||
latin = []
|
||
for i in range(size):
|
||
row = [(i + j) % size + 1 for j in range(size)]
|
||
latin.append(row)
|
||
random.shuffle(latin)
|
||
return latin
|
||
|
||
@staticmethod
|
||
def _generate_thermometer_path(solution, start_row, start_col):
|
||
path = [(start_row, start_col)]
|
||
current_value = solution[start_row][start_col]
|
||
size = len(solution)
|
||
|
||
while True:
|
||
last_row, last_col = path[-1]
|
||
neighbors = []
|
||
for dr, dc in [(-1,0), (1,0), (0,-1), (0,1)]:
|
||
nr, nc = last_row + dr, last_col + dc
|
||
if 0 <= nr < size and 0 <= nc < size and (nr, nc) not in path:
|
||
next_value = solution[nr][nc]
|
||
if next_value > current_value:
|
||
neighbors.append((nr, nc, next_value))
|
||
if not neighbors:
|
||
break
|
||
nr, nc, nv = random.choice(neighbors)
|
||
path.append((nr, nc))
|
||
current_value = nv
|
||
|
||
return path if len(path) >= 2 else None
|
||
|
||
@staticmethod
|
||
def prompt_func(question_case) -> str:
|
||
def format_coord(coord):
|
||
r, c = coord
|
||
return f"行{r+1}列{c+1}"
|
||
|
||
size = question_case['size']
|
||
thermometers = question_case['thermometers']
|
||
enforce_latin = question_case['enforce_latin']
|
||
|
||
rules = [
|
||
f"1. 在{size}x{size}的网格中填入1到{size}的数字。",
|
||
"2. 每个温度计的路径必须从灯泡(●)开始严格递增。",
|
||
f"3. 每{'行和列必须包含不重复的1到{size}' if enforce_latin else '行和列允许重复但需满足温度计约束'}。"
|
||
]
|
||
|
||
thermo_desc = []
|
||
for i, thermo in enumerate(thermometers, 1):
|
||
path = thermo['path']
|
||
bulb = format_coord(thermo['bulb'])
|
||
tip = format_coord(path[-1])
|
||
path_str = " → ".join(format_coord(p) for p in path)
|
||
thermo_desc.append(f"温度计{i}: 从{bulb}到{tip}, 路径: {path_str}")
|
||
|
||
return (
|
||
"解决以下温度计谜题：\n\n" +
|
||
"\n".join(rules) + "\n\n" +
|
||
"温度计列表：\n" + "\n".join(thermo_desc) + "\n\n" +
|
||
"将答案按行排列，每行数字用空格分隔，置于[answer]和[/answer]之间。示例：\n" +
|
||
"[answer]\n1 2 3\n2 3 1\n3 1 2\n[/answer]"
|
||
)
|
||
|
||
@staticmethod
|
||
def extract_output(output):
|
||
answer_blocks = re.findall(r'\[answer\](.*?)\[\/answer\]', output, re.DOTALL)
|
||
if not answer_blocks:
|
||
return None
|
||
last_answer = answer_blocks[-1].strip()
|
||
rows = [line.strip() for line in last_answer.split('\n') if line.strip()]
|
||
try:
|
||
grid = [list(map(int, row.split())) for row in rows]
|
||
if all(len(row) == len(grid[0]) for row in grid) and len(grid) == len(grid[0]):
|
||
return grid
|
||
except ValueError:
|
||
pass
|
||
return None
|
||
|
||
@classmethod
|
||
def _verify_correction(cls, solution, identity):
|
||
size = identity['size']
|
||
enforce_latin = identity.get('enforce_latin', True)
|
||
thermometers = identity['thermometers']
|
||
|
||
if len(solution) != size or any(len(row) != size for row in solution):
|
||
return False
|
||
if any(not (1 <= num <= size) for row in solution for num in row):
|
||
return False
|
||
|
||
if enforce_latin:
|
||
expected = list(range(1, size+1))
|
||
for row in solution:
|
||
if sorted(row) != expected:
|
||
return False
|
||
for col in range(size):
|
||
if sorted(row[col] for row in solution) != expected:
|
||
return False
|
||
|
||
for thermo in thermometers:
|
||
path = thermo['path']
|
||
values = [solution[r][c] for (r, c) in path]
|
||
if any(values[i] >= values[i+1] for i in range(len(values)-1)):
|
||
return False
|
||
|
||
return True
|