mirror of
https://github.com/InternLM/InternBootcamp.git
synced 2026-04-19 12:58:04 +00:00
init-commit
This commit is contained in:
commit
18a552597a
3461 changed files with 1150579 additions and 0 deletions
336
internbootcamp/bootcamp/kor_operation_unicode25bd/kor_operation_unicode25bd.py
Executable file
336
internbootcamp/bootcamp/kor_operation_unicode25bd/kor_operation_unicode25bd.py
Executable file
|
|
@ -0,0 +1,336 @@
|
|||
"""# 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
|
||||
from bootcamp import Basebootcamp
|
||||
|
||||
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
|
||||
注意:请参照提供的谜题描述进行复述,规则应当描述详细,包括任务背景、具体任务操作规则、对题目格式和答案格式的含义介绍等,
|
||||
|
||||
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(如双括号)内,例如[[your answer here]]
|
||||
\"\"\"
|
||||
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
|
||||
f▽g=f(x) \quad+g''(x) \quad.Example questions are as follows:
|
||||
|
||||
<example 0>
|
||||
f(x)=x^2, g(x)=sin(x), compute f▽g.
|
||||
Please provide your answer in LaTeX format.
|
||||
Wrap the final answer in double square brackets, like this: [[your answer]].
|
||||
</example 0>
|
||||
|
||||
<example 1>
|
||||
f(x)=e^x, g(x)=ln(x) find the value of f▽g when x=1.
|
||||
Please wrap the answer in double square brackets, like this: [[your answer]].
|
||||
</example 1>
|
||||
|
||||
<example 2>
|
||||
f(x)=cos(x), g(x)=x^3 find the value of f▽g when x=0.
|
||||
Please ensure the answer is a single number and wrap it in double square brackets, like this: [[your answer]].
|
||||
</example 2>
|
||||
|
||||
<example 3>
|
||||
f(x)=ln(x), g(x)=e^x, find the value of f▽g when x=1.
|
||||
Please wrap the answer in double square brackets, like this: [[your answer]].
|
||||
</example 3>
|
||||
|
||||
<example 4>
|
||||
f(x)=\sqrt{x},g(x)=cos(x),compute f▽g.
|
||||
Please provide your answer in LaTeX format.
|
||||
Wrap the final answer in double square brackets, like this: [[your answer]].
|
||||
</example 4>
|
||||
|
||||
<example 5>
|
||||
f(x)=sin(x), g(x)=ln(x), compute f▽g.
|
||||
Please provide your answer in LaTeX format.
|
||||
Wrap the final answer in double square brackets, like this: [[your answer]].
|
||||
</example 5>
|
||||
|
||||
<example 6>
|
||||
f(x)=e^x,g(x)=sin(x),find the value of f▽g when x=0.
|
||||
Please ensure the answer is a single number and wrap it in double square brackets, like this: [[your answer]].
|
||||
</example 6>
|
||||
|
||||
<example 7>
|
||||
f(x)=ln(x), g(x)=x^2,find the value of f▽g when x=1.
|
||||
Please ensure the answer is a single number and wrap it in double square brackets, like this: [[your answer]].
|
||||
</example 7>
|
||||
|
||||
<example 8>
|
||||
f(x)=tan(x), g(x)=x,find the value of f▽g when x=π/4.
|
||||
Please ensure the answer is a single number and wrap it in double square brackets, like this: [[your answer]].
|
||||
</example 8>
|
||||
|
||||
<example 9>
|
||||
f(x)=x^3,g(x)=e^x,find the value of f▽g when x=1.
|
||||
Please wrap the answer in double square brackets, like this: [[your answer]].
|
||||
</example 9>
|
||||
|
||||
|
||||
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 sympy as sp
|
||||
import random
|
||||
|
||||
class KorOperationUnicode25bdbootcamp(Basebootcamp):
|
||||
def __init__(self, x_value_prob=0.3, numeric_precision=4, **kwargs):
|
||||
super().__init__()
|
||||
self.x_value_prob = x_value_prob # 数值问题的概率
|
||||
self.numeric_precision = numeric_precision
|
||||
|
||||
def safe_generate(self, func_type):
|
||||
"""生成定义域安全的函数表达式"""
|
||||
x = sp.symbols('x')
|
||||
|
||||
# 控制函数生成范围
|
||||
if func_type == 'logarithm':
|
||||
base = random.choice([sp.E, 10])
|
||||
return random.randint(1,3)*sp.log(base**random.randint(1,3)*x)
|
||||
elif func_type == 'polynomial':
|
||||
return sum(random.randint(1,3)*x**i for i in range(3))
|
||||
elif func_type == 'trigonometric':
|
||||
choice = random.choice([sp.sin, sp.cos])
|
||||
return random.randint(1,3)*choice(random.randint(1,3)*x)
|
||||
else: # 指数函数
|
||||
return random.randint(1,3)*sp.exp(random.randint(1,3)*x)
|
||||
|
||||
def generate_case_components(self):
|
||||
"""生成合法的问题组件"""
|
||||
x = sp.symbols('x')
|
||||
for _ in range(100): # 尝试次数限制
|
||||
# 控制函数类型组合
|
||||
f_type = random.choice(['polynomial', 'trigonometric', 'exponential'])
|
||||
g_type = random.choice(['polynomial', 'trigonometric', 'logarithm'])
|
||||
|
||||
f_expr = self.safe_generate(f_type)
|
||||
g_expr = self.safe_generate(g_type)
|
||||
|
||||
# 计算二阶导数
|
||||
try:
|
||||
g_double_prime = sp.diff(g_expr, x, 2)
|
||||
except:
|
||||
continue
|
||||
|
||||
# 生成合法的x值
|
||||
x_value = self.find_valid_x(f_expr, g_expr)
|
||||
if x_value is None:
|
||||
continue
|
||||
|
||||
return f_expr, g_expr, g_double_prime, x_value
|
||||
|
||||
# 保底返回
|
||||
return x, sp.sin(x), 0, 1.0
|
||||
|
||||
def find_valid_x(self, f_expr, g_expr):
|
||||
"""寻找满足所有条件的x值"""
|
||||
x = sp.symbols('x')
|
||||
for _ in range(100):
|
||||
# 根据函数类型调整取值范围
|
||||
if any(func.has(sp.log(x)) for func in [g_expr]):
|
||||
x_candidate = random.uniform(0.1, 5)
|
||||
else:
|
||||
x_candidate = random.uniform(-3, 3)
|
||||
|
||||
try:
|
||||
f_expr.subs(x, x_candidate)
|
||||
g_expr.subs(x, x_candidate)
|
||||
return round(x_candidate, 2)
|
||||
except:
|
||||
continue
|
||||
return None
|
||||
|
||||
def case_generator(self):
|
||||
f_expr, g_expr, g_double_prime, x_value = self.generate_case_components()
|
||||
|
||||
# 生成两种问题类型
|
||||
is_numeric = random.random() < self.x_value_prob
|
||||
expected_str = None
|
||||
expected_num = None
|
||||
|
||||
x = sp.symbols('x')
|
||||
correct_expr = f_expr + g_double_prime
|
||||
|
||||
if is_numeric:
|
||||
# 数值计算
|
||||
numeric_value = correct_expr.subs(x, x_value).evalf()
|
||||
expected_num = round(float(numeric_value), self.numeric_precision)
|
||||
else:
|
||||
# 符号表达式处理
|
||||
expected_str = sp.latex(correct_expr.simplify())
|
||||
|
||||
return {
|
||||
'f_latex': sp.latex(f_expr),
|
||||
'g_latex': sp.latex(g_expr),
|
||||
'x_value': x_value if is_numeric else None,
|
||||
'expected_num': expected_num,
|
||||
'expected_str': expected_str,
|
||||
'precision': self.numeric_precision,
|
||||
'is_numeric': is_numeric
|
||||
}
|
||||
|
||||
@staticmethod
|
||||
def prompt_func(question_case) -> str:
|
||||
problem = [
|
||||
"Solve the differential operator problem:",
|
||||
"Given:",
|
||||
f"f(x) = {question_case['f_latex']}",
|
||||
f"g(x) = {question_case['g_latex']}",
|
||||
"Compute: f▽g = f(x) + g''(x)"
|
||||
]
|
||||
|
||||
if question_case['is_numeric']:
|
||||
problem.append(f"at x = {question_case['x_value']}")
|
||||
problem.append(f"Provide a numerical value rounded to {question_case['precision']} decimal places.")
|
||||
else:
|
||||
problem.append("Provide the result as a LaTeX mathematical expression.")
|
||||
|
||||
problem.append("Format your answer within double square brackets: [[answer]]")
|
||||
return '\n'.join(problem)
|
||||
|
||||
@staticmethod
|
||||
def extract_output(output):
|
||||
import re
|
||||
matches = re.findall(r'\[\[(.*?)\]\]', output)
|
||||
return matches[-1].strip() if matches else None
|
||||
|
||||
@classmethod
|
||||
def _verify_correction(cls, solution, identity):
|
||||
try:
|
||||
if identity['is_numeric']:
|
||||
# 数值验证
|
||||
user_val = round(float(solution), identity['precision'])
|
||||
return abs(user_val - identity['expected_num']) < 1e-6
|
||||
else:
|
||||
# 符号验证
|
||||
x = sp.symbols('x')
|
||||
user_expr = sp.sympify(solution, evaluate=False)
|
||||
expected_expr = sp.sympify(identity['expected_str'], evaluate=False)
|
||||
return sp.simplify(user_expr - expected_expr) == 0
|
||||
except:
|
||||
return False
|
||||
Loading…
Add table
Add a link
Reference in a new issue