mirror of
https://github.com/InternLM/InternBootcamp.git
synced 2026-04-28 17:29:37 +00:00
246 lines
9.5 KiB
Python
Executable file
246 lines
9.5 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
|
||
|
||
The Aquarium puzzle is solved by determining water levels for each aquarium region in a grid, adhering to the following rules:
|
||
|
||
1. **Grid Structure**: The grid is divided into contiguous regions (aquariums) by thick borders. Each cell belongs to exactly one aquarium.
|
||
|
||
2. **Water Levels**: Each aquarium must be filled with water up to a consistent horizontal level. This level is a specific row number chosen such that:
|
||
- Every column within the aquarium contains cells up to at least this row (i.e., the level cannot exceed the shortest column height in the aquarium).
|
||
- All cells in the aquarium’s columns from the bottom row up to the chosen level are filled. Cells above this level in the aquarium remain empty.
|
||
|
||
3. **Row and Column Clues**:
|
||
- Numbers on the right side of each row indicate the total number of filled cells required in that row across all aquariums.
|
||
- Numbers at the bottom/top of each column indicate the total number of filled cells required in that column across all aquariums.
|
||
|
||
4. **Objective**: Fill cells to satisfy all row/column numerical clues while ensuring each aquarium’s water level is uniformly applied to its columns.
|
||
|
||
|
||
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
|
||
from typing import Dict, List, Optional
|
||
|
||
class Aquariumbootcamp(Basebootcamp):
|
||
def __init__(self, grid_rows: int = 5, grid_cols: int = 5):
|
||
self.grid_rows = grid_rows
|
||
self.grid_cols = grid_cols
|
||
|
||
def case_generator(self) -> Dict:
|
||
# Generate regions where each column is a separate aquarium
|
||
cols = self.grid_cols
|
||
rows = self.grid_rows
|
||
regions = []
|
||
for r in range(rows):
|
||
regions.append([c for c in range(cols)])
|
||
|
||
# Generate water levels for each column (aquarium)
|
||
k = [random.randint(0, rows - 1) for _ in range(cols)]
|
||
|
||
# Compute row clues: number of filled cells per row
|
||
row_clues = []
|
||
for r in range(rows):
|
||
count = sum(1 for c in range(cols) if k[c] >= r)
|
||
row_clues.append(count)
|
||
|
||
# Column clues are k[i] + 1
|
||
col_clues = [ki + 1 for ki in k]
|
||
|
||
return {
|
||
'regions': regions,
|
||
'row_clues': row_clues,
|
||
'col_clues': col_clues,
|
||
}
|
||
|
||
@staticmethod
|
||
def prompt_func(question_case: Dict) -> str:
|
||
rows = len(question_case['regions'])
|
||
cols = len(question_case['regions'][0]) if rows > 0 else 0
|
||
regions_table = '\n'.join([f"Row {i}: {' '.join(map(str, row))}" for i, row in enumerate(question_case['regions'])])
|
||
row_clues = question_case['row_clues']
|
||
col_clues = question_case['col_clues']
|
||
|
||
prompt = f"""You are to solve an Aquarium puzzle. The puzzle is played on a grid divided into aquarium regions. Each aquarium must be filled up to a horizontal level such that all its columns are filled to the same level. Here are the details:
|
||
|
||
- The grid has {rows} rows and {cols} columns.
|
||
|
||
- Aquarium regions are as follows (each number represents the aquarium ID for that cell):
|
||
{regions_table}
|
||
|
||
- Each row has a clue on the right indicating the total filled cells in that row. The row clues are: {row_clues}.
|
||
|
||
- Each column has a clue at the bottom indicating the total filled cells in that column. The column clues are: {col_clues}.
|
||
|
||
Your task is to determine the water level for each aquarium. The water level is the highest row number filled (0-based from the bottom). Each aquarium's water level must be such that all its columns are filled up to this level.
|
||
|
||
Provide your answer as a list of integers in column order (from left to right), where each integer is the water level for the corresponding column's aquarium. Enclose your answer within [answer] and [/answer]. For example, if the solution is levels 2, 1, 0 for columns 0, 1, 2, write:
|
||
[answer]2 1 0[/answer]"""
|
||
return prompt
|
||
|
||
@staticmethod
|
||
def extract_output(output: str) -> Optional[List[int]]:
|
||
# Find all answer blocks
|
||
matches = re.findall(r'\[answer\](.*?)\[/answer\]', output, re.DOTALL)
|
||
if not matches:
|
||
return None
|
||
# Take the last match
|
||
last_match = matches[-1].strip()
|
||
try:
|
||
solution = list(map(int, last_match.split()))
|
||
return solution
|
||
except:
|
||
return None
|
||
|
||
@classmethod
|
||
def _verify_correction(cls, solution: List[int], identity: Dict) -> bool:
|
||
cols = len(identity['col_clues'])
|
||
rows = len(identity['row_clues'])
|
||
# Check solution length matches columns
|
||
if len(solution) != cols:
|
||
return False
|
||
# Check each column's solution matches column clue
|
||
for c in range(cols):
|
||
if solution[c] + 1 != identity['col_clues'][c]:
|
||
return False
|
||
# Check each row's filled count matches row clue
|
||
for r in range(rows):
|
||
expected = identity['row_clues'][r]
|
||
actual = sum(1 for c in range(cols) if solution[c] >= r)
|
||
if actual != expected:
|
||
return False
|
||
return True
|