mirror of
https://github.com/InternLM/InternBootcamp.git
synced 2026-04-19 12:58:04 +00:00
244 lines
8.8 KiB
Python
Executable file
244 lines
8.8 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
|
||
|
||
Kakurasu is a logic puzzle played on a rectangular grid (typically N×N). Each cell in the grid can be either shaded or unshaded. The puzzle provides target values for each row and column, and the goal is to shade cells such that:
|
||
|
||
1. **Row Constraints**: For every row, the sum of the *column indices* of the shaded cells in that row equals the target value specified for that row.
|
||
- Example: If a row's target is 7, you might shade cells in columns 3 and 4 (since 3 + 4 = 7).
|
||
|
||
2. **Column Constraints**: For every column, the sum of the *row indices* of the shaded cells in that column equals the target value specified for that column.
|
||
- Example: If a column's target is 5, you might shade cells in rows 2 and 3 (since 2 + 3 = 5).
|
||
|
||
3. **Unique Contributions**: A shaded cell at position (row *i*, column *j*) contributes its *column index* **j** to its row's sum and its *row index* **i** to its column's sum. These dual contributions must satisfy both the row and column targets simultaneously.
|
||
|
||
4. **No Overlapping Rules**: Unlike Sudoku, there are no region constraints—only row and column sums matter. However, cells cannot be \"partially\" shaded; they are either fully shaded or unshaded.
|
||
|
||
The puzzle is solved when all row and column targets are satisfied without contradiction.
|
||
|
||
|
||
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 random
|
||
import re
|
||
import ast
|
||
|
||
class Kakurasubootcamp(Basebootcamp):
|
||
def __init__(self, n=5):
|
||
self.n = n
|
||
|
||
def case_generator(self):
|
||
n = self.n
|
||
grid = [[random.choice([0, 1]) for _ in range(n)] for _ in range(n)]
|
||
|
||
row_targets = []
|
||
for i in range(n):
|
||
total = sum((j + 1) * cell for j, cell in enumerate(grid[i]))
|
||
row_targets.append(total)
|
||
|
||
col_targets = []
|
||
for j in range(n):
|
||
total = sum((i + 1) * grid[i][j] for i in range(n))
|
||
col_targets.append(total)
|
||
|
||
return {
|
||
'n': n,
|
||
'row_targets': row_targets,
|
||
'col_targets': col_targets
|
||
}
|
||
|
||
@staticmethod
|
||
def prompt_func(question_case) -> str:
|
||
n = question_case['n']
|
||
row_targets = question_case['row_targets']
|
||
col_targets = question_case['col_targets']
|
||
return f"""你正在解决一个Kakurasu谜题。这是一个{n}x{n}的网格谜题，目标是根据行和列的约束条件涂黑单元格。
|
||
|
||
规则：
|
||
1. 每行中被涂黑单元格的列索引（从左到右为1到{n}）之和等于该行的目标值。
|
||
2. 每列中被涂黑单元格的行索引（从上到下为1到{n}）之和等于该列的目标值。
|
||
3. 每个单元格必须明确涂黑（1）或未涂黑（0）。
|
||
|
||
当前谜题的行目标值（从上到下）：{row_targets}
|
||
当前谜题的列目标值（从左到右）：{col_targets}
|
||
|
||
请将你的解答格式化为{n}x{n}的二维数组，其中1表示涂黑，0表示未涂黑，并用[answer]和[/answer]标签包裹。例如：
|
||
[answer]
|
||
[[1, 0, 0], [0, 1, 0], [0, 0, 1]]
|
||
[/answer]
|
||
"""
|
||
|
||
@staticmethod
|
||
def extract_output(output):
|
||
matches = re.findall(r'\[answer\](.*?)\[/answer\]', output, re.DOTALL)
|
||
if not matches:
|
||
return None
|
||
last_match = matches[-1].strip()
|
||
try:
|
||
solution = ast.literal_eval(last_match)
|
||
return solution
|
||
except (SyntaxError, ValueError):
|
||
return None
|
||
|
||
@classmethod
|
||
def _verify_correction(cls, solution, identity):
|
||
n = identity['n']
|
||
row_targets = identity['row_targets']
|
||
col_targets = identity['col_targets']
|
||
|
||
# 验证答案结构
|
||
if not isinstance(solution, list) or len(solution) != n:
|
||
return False
|
||
for row in solution:
|
||
if not isinstance(row, list) or len(row) != n:
|
||
return False
|
||
for cell in row:
|
||
if cell not in (0, 1):
|
||
return False
|
||
|
||
# 验证行约束
|
||
for i in range(n):
|
||
row_sum = sum((j + 1) * cell for j, cell in enumerate(solution[i]))
|
||
if row_sum != row_targets[i]:
|
||
return False
|
||
|
||
# 验证列约束
|
||
for j in range(n):
|
||
col_sum = sum((i + 1) * solution[i][j] for i in range(n))
|
||
if col_sum != col_targets[j]:
|
||
return False
|
||
|
||
return True
|