Merge branch 'hf'

README.md

@@ -28,12 +28,26 @@ Welcome to LLM Riddles! This is a game of wits and courage with language models.
 
 ## :space_invader: How to Play
 We provide an online version for players to directly access and try out. 
-- [ChatGPT + English(w/o key)](https://huggingface.co/spaces/OpenDILabCommunity/LLMRiddlesChatGPTEN)
-- [ChatGPT + Chinese(w/o key)](https://huggingface.co/spaces/OpenDILabCommunity/LLMRiddlesChatGPTCN)
-- [Mistral + English(w/ key)](https://4521e4d138d3779498.gradio.live)
-- [ChatGPT + Chinese(w/ key)](http://llmriddles.opendilab.net/)
+- [Hugging Face][ChatGPT + English(w/o key)](https://huggingface.co/spaces/OpenDILabCommunity/LLMRiddlesChatGPTEN)
+- [Hugging Face][ChatGPT + Chinese(w/o key)](https://huggingface.co/spaces/OpenDILabCommunity/LLMRiddlesChatGPTCN)
+- [Hugging Face][ChatGLM + Chinese(w/ key)](https://huggingface.co/spaces/OpenDILabCommunity/LLMRiddlesChatGLMCN)
+- [OpenXLab][ChatGPT + Chinese(w/o key)](https://openxlab.org.cn/apps/detail/OpenDILab/LLMRiddlesChatGPTCN)
+- [OpenXLab][ChatGLM + Chinese(w/ key)](https://openxlab.org.cn/apps/detail/OpenDILab/LLMRiddlesChatGLMCN)
+- [OpenXLab][ChatGLM + English(w/ key)](https://openxlab.org.cn/apps/detail/OpenDILab/LLMRiddlesChatGLMEN)
+- [Private Server][Mistral + English(w/ key)](https://d9b451a97791dd8ef3.gradio.live)
+- [Private Server][ChatGPT + Chinese(w/ key)](http://llmriddles.opendilab.net/)
 
 Local deployment can be done in the following ways:
+## Installation
+### Use ChatGPT / ChatGLM API
+```shell
+pip3 install -r requirements.txt
+```
+### Deploy Mistral-7B-Instruct-v0.1 for local inference
+```shell
+pip3 install -r requirements-dev.txt
+```
+## Launch
 ### ChatGPT + Chinese
 ```shell
 QUESTION_LANG=cn QUESTION_LLM='chatgpt' QUESTION_LLM_KEY=<your API key> python3 -u app.py
@@ -42,6 +56,14 @@ QUESTION_LANG=cn QUESTION_LLM='chatgpt' QUESTION_LLM_KEY=<your API key> python3
 ```shell
 QUESTION_LANG=en QUESTION_LLM='chatgpt' QUESTION_LLM_KEY=<your API key> python3 -u app.py
 ```
+### ChatGLM + Chinese
+```shell
+QUESTION_LANG=cn QUESTION_LLM='chatglm' QUESTION_LLM_KEY=<your API key> python3 -u app.py
+```
+### ChatGLM + English
+```shell
+QUESTION_LANG=en QUESTION_LLM='chatglm' QUESTION_LLM_KEY=<your API key> python3 -u app.py
+```
 ### Mistral-7B-Instruct-v0.1 + English
 ```shell
 QUESTION_LANG=en QUESTION_LLM='mistral-7b' python3 -u app.py
@@ -56,7 +78,7 @@ The question format should include the following points:
 - Pull Request title, example: feature(username): Chapter X-Level Design
 - The ID you want to be mentioned
 - Modify the corresponding chapter question files
-- Modification of init.py
+- Modification of \__init__.py
 
 For a complete example, please refer to: [Submit your own level design](https://github.com/opendilab/LLMRiddles/pull/6)
 
@@ -66,10 +88,11 @@ For a complete example, please refer to: [Submit your own level design](https://
 - [x] Online trial link
 - [x] Hugging Face Space link
 - [x] Support Mistral-7B（English version）
-- [ ] Support ChatGLM（Chinese version）
+- [x] Support ChatGLM（Chinese and English version）
 - [ ] Support Baichuan2-7B（Chinese version）
 - [ ] Support LLaMA2-7B（English version）
 - [ ] LLM inference speed optimization
+- [ ] More question levels and solution blogs
 
 ## :speech_balloon: Feedback and Contribution
 - [Start an Issue](https://github.com/opendilab/CodeMorpheus/issues/new/choose) on GitHub
@@ -80,6 +103,7 @@ For a complete example, please refer to: [Submit your own level design](https://
 ## :star2: Special Thanks
 - Thanks to [Haoqiang Fan](https://www.zhihu.com/people/haoqiang-fan) for his original idea and title, which provided inspiration and motivation for the development and expansion of this project.
 - Thanks to [HuggingFace](https://huggingface.co) for supporting and assisting the game.
+- Thanks to [ChatGLM](https://chatglm.cn) for supporting and assisting the game, especially sufficient inference token support.
 - Thanks to [LLM Riddles contributors](https://github.com/opendilab/LLMRiddles/graphs/contributors) for their implementation and support.
 
 ## :label: License

README_zh.md

@@ -15,12 +15,26 @@
 
 ## :space_invader: 如何试玩
 我们提供了在线版本以供玩家直接访问试玩:
-- [ChatGPT + 英文(需配置api key)](https://huggingface.co/spaces/OpenDILabCommunity/LLMRiddlesChatGPTEN)
-- [ChatGPT + 中文(需配置api key)](https://huggingface.co/spaces/OpenDILabCommunity/LLMRiddlesChatGPTCN)
-- [Mistral + 英文(已预设api key)](https://4521e4d138d3779498.gradio.live)
-- [ChatGPT + 中文(已预设api key)](http://llmriddles.opendilab.net/)
+- [Hugging Face][ChatGPT + 英文(需配置api key)](https://huggingface.co/spaces/OpenDILabCommunity/LLMRiddlesChatGPTEN)
+- [Hugging Face][ChatGPT + 中文(需配置api key)](https://huggingface.co/spaces/OpenDILabCommunity/LLMRiddlesChatGPTCN)
+- [Hugging Face][ChatGLM + 中文(已预设api key)](https://huggingface.co/spaces/OpenDILabCommunity/LLMRiddlesChatGLMCN)
+- [OpenXLab][ChatGPT + 中文(需配置api key)](https://openxlab.org.cn/apps/detail/OpenDILab/LLMRiddlesChatGPTCN)
+- [OpenXLab][ChatGLM + 中文(已预设api key)](https://openxlab.org.cn/apps/detail/OpenDILab/LLMRiddlesChatGLMCN)
+- [OpenXLab][ChatGLM + 英文(已预设api key)](https://openxlab.org.cn/apps/detail/OpenDILab/LLMRiddlesChatGLMEN)
+- [Private Server][Mistral + 英文(已预设api key)](https://d9b451a97791dd8ef3.gradio.live)
+- [Private Server][ChatGPT + 中文(已预设api key)](http://llmriddles.opendilab.net/)
 
 本地部署可以通过以下方式：
+## 安装
+### ChatGPT 或 ChatGLM API
+```shell
+pip3 install -r requirements.txt
+```
+### Mistral-7B-Instruct-v0.1 本地推理
+```shell
+pip3 install -r requirements-dev.txt
+```
+## 启动
 ### ChatGPT + 中文
 ```shell
 QUESTION_LANG=cn QUESTION_LLM='chatgpt' QUESTION_LLM_KEY=<your API key> python3 -u app.py
@@ -29,13 +43,17 @@ QUESTION_LANG=cn QUESTION_LLM='chatgpt' QUESTION_LLM_KEY=<your API key> python3
 ```shell
 QUESTION_LANG=en QUESTION_LLM='chatgpt' QUESTION_LLM_KEY=<your API key> python3 -u app.py
 ```
-### LLaMA2-7b + 中文
+### ChatGLM + 中文
+```shell
+QUESTION_LANG=cn QUESTION_LLM='chatglm' QUESTION_LLM_KEY=<your API key> python3 -u app.py
+```
+### ChatGLM + 英文
 ```shell
-QUESTION_LANG=cn QUESTION_LLM='llama2-7b' python3 -u app.py
+QUESTION_LANG=en QUESTION_LLM='chatglm' QUESTION_LLM_KEY=<your API key> python3 -u app.py
 ```
-### LLaMA2-7b + 英文
+### Mistral-7B-Instruct-v0.1 + 英文
 ```shell
-QUESTION_LANG=en QUESTION_LLM='llama2-7b' python3 -u app.py
+QUESTION_LANG=en QUESTION_LLM='mistral-7b' python3 -u app.py
 ```
 ## :technologist: 为什么制作这个游戏
 
@@ -48,7 +66,7 @@ QUESTION_LANG=en QUESTION_LLM='llama2-7b' python3 -u app.py
 - Pull Request标题，示例：feature(username): 章节X-关卡设计
 - 希望被提及的ID
 - 对应章节问题文件的修改
-- init.py的修改
+- \__init__.py的修改
 
 完整示例请参考：[提交属于自己的关卡设计](https://github.com/opendilab/LLMRiddles/pull/6)
 
@@ -58,10 +76,11 @@ QUESTION_LANG=en QUESTION_LLM='llama2-7b' python3 -u app.py
 - [x] 在线试玩链接
 - [x] Hugging Face Space 链接
 - [x] 支持Mistral-7B（英文）
-- [ ] 支持ChatGLM（中文）
+- [x] 支持ChatGLM（中文和英文）
 - [ ] 支持Baichuan2-7B（中文）
 - [ ] 支持LLaMA2-7B（英文）
 - [ ] LLM 推理速度优化
+- [ ] 更多题目和题解
 
 ## :speech_balloon: 反馈问题 & 提出建议
 - 在 GitHub 上[发起 Issue](https://github.com/opendilab/CodeMorpheus/issues/new/choose)
@@ -72,6 +91,7 @@ QUESTION_LANG=en QUESTION_LLM='llama2-7b' python3 -u app.py
 ## :star2: Special Thanks
 - 感谢 [Haoqiang Fan](https://www.zhihu.com/people/haoqiang-fan) 的原始创意和题目，为本项目的开发和扩展提供了灵感与动力。
 - 感谢 [HuggingFace](https://huggingface.co) 对游戏的支持与协助。
+- 感谢 [ChatGLM](https://chatglm.cn/) 对游戏的支持与协助，特别是供在线预览版使用的足量 tokens。
 - 感谢 [LLM Riddles contributors](https://github.com/opendilab/LLMRiddles/graphs/contributors) 的实现与支持。
 
 ## :label: License

app.py

@@ -13,7 +13,7 @@ _QUESTIONS = list_ordered_questions()
 _LANG = os.environ.get('QUESTION_LANG', 'cn')
 assert _LANG in ['cn', 'en'], _LANG
 _LLM = os.environ.get('QUESTION_LLM', 'chatgpt')
-assert _LLM in ['chatgpt', 'mistral-7b'], _LLM
+assert _LLM in ['chatgpt', 'chatglm', 'mistral-7b'], _LLM
 _LLM_KEY = os.environ.get('QUESTION_LLM_KEY', None)
 _DEBUG = os.environ.get('DEBUG', 'false').lower() == 'true'
 
@@ -49,6 +49,7 @@ if _LANG == "cn":
         <img src="https://raw.githubusercontent.com/opendilab/LLMRiddles/main/llmriddles/assets/banner.svg" width="80%" height="20%" alt="Banner Image">
     </div>
     <h2 style="text-align: center; color: black;"><a href="https://github.com/OpenDILab/LLMRiddles"> 🎭LLM Riddles：完蛋！我被 LLM 拿捏了</a></h2>
+    <strong><h5 align="center"> 其他在线示例：中文在线试玩版本<a href="https://openxlab.org.cn/apps/detail/OpenDILab/LLMRiddlesChatGLMCN">（OpenXLab）</a> | 中文在线试玩版本<a href="https://huggingface.co/spaces/OpenDILabCommunity/LLMRiddlesChatGLMCN">（Hugging Face）</a> <h5></strong>
     <strong><h5 align="center"> 更多不同语言模型的在线试玩 demo 可以访问 GitHub<a href="https://github.com/OpenDILab/LLMRiddles">源代码仓库</a>获取<h5></strong>
     <h5 align="center"> 如果你喜欢这个项目，请给我们在 GitHub 点个 star ✨ <a href="https://github.com/OpenDILab/LLMRiddles"> 代码仓库传送门 </a> 。我们将会持续保持更新。再次感谢游戏<a href="https://www.zhihu.com/people/haoqiang-fan"> 原作者 </a>的奇思妙想！  </h5>
     <strong><h5 align="center">注意：算法模型的输出可能包含一定的随机性。相关结果不代表任何开发者和相关 AI 服务的态度和意见。本项目开发者不对生成结果作任何保证，仅供娱乐。<h5></strong>
@@ -105,12 +106,14 @@ else:
 
 
 def _need_api_key():
-    return _LLM == 'chatgpt' and _LLM_KEY is None
+    return (_LLM == 'chatgpt' or _LLM == 'chatglm') and _LLM_KEY is None
 
 
 def _get_api_key_cfgs(api_key):
     if _LLM == 'chatgpt':
         return {'api_key': api_key}
+    elif _LLM == 'chatglm':
+        return {'api_key': api_key}
     else:
         return {}
 

llmriddles/llms/__init__.py

@@ -1,3 +1,4 @@
 from .base import register_llm, get_llm_fn
 from .chatgpt import ask_chatgpt
+from .chatglm import ask_chatglm
 from .mistral import ask_mistral_7b_instruct

llmriddles/llms/chatglm.py

@@ -0,0 +1,26 @@
+import zhipuai
+from .base import register_llm
+
+
+def ask_chatglm(message: str, api_key: str):
+    zhipuai.api_key = api_key
+
+    response = zhipuai.model_api.invoke(
+        model="chatglm_turbo",
+        prompt=[{
+            "role": "user",
+            "content": message
+        }],
+        top_p=0.7,
+        temperature=0.9,
+    )
+
+    response_msg = response['data']['choices'][0]['content']
+    # strip the front and end '"'
+    if len(response_msg) >= 2:
+        response_msg = response_msg[1:-1]
+
+    return response_msg
+
+
+register_llm('chatglm', ask_chatglm)

llmriddles/questions/level4.py

@@ -82,12 +82,12 @@ def _checker_3(question_text: str, user_text: str, answer_text: str, lang: str):
     _ = question_text, lang
     answer_text = answer_text.strip()
     user_text = user_text.strip()
-    if not check_if_is_number(question_text):
+    if not check_if_is_number(user_text):
         return False, "问题应该是一个正整数" if lang == 'cn' else 'Question should be a positive integer.'
-    elif int(question_text) == 1:
+    elif int(user_text) == 1:
         return False, "问题应该是一个大于1的正整数" if lang == 'cn' else 'Question should be a positive integer greater than 1.'
-    elif int(question_text) - 1 not in get_all_numbers_in_a_sentence(answer_text) or int(
-            question_text) + 1 not in get_all_numbers_in_a_sentence(answer_text):
+    elif int(user_text) - 1 not in get_all_numbers_in_a_sentence(answer_text) or int(
+            user_text) + 1 not in get_all_numbers_in_a_sentence(answer_text):
         return False, "回答中应该包含一个与问题相差1的数字" if lang == 'cn' else 'Answer should contain a number that is exactly 1 different from the question.'
     else:
         return True, None

llmriddles/questions/level5.py

@@ -47,7 +47,11 @@ def _checker_1(question_text: str, user_text: str, answer_text: str, lang: str):
     if count_words(user_text) < 10:
         return False, "用户的问题长度应该至少10个字" if lang == 'cn' else 'Question should be no less than 10 words.'
 
-    if answer_text == user_text:
+    if lang == 'en':
+        answer_text = answer_text.split()
+        user_text = user_text.split()
+    
+    if not any(each in answer_text for each in user_text):
         return True, None
     else:
         return False, None

requirements-dev.txt

@@ -0,0 +1,11 @@
+hbutils>=0.9.1
+tqdm
+requests>=2.20
+gradio==4.1.1
+openai>=1
+zhipuai
+sympy
+flask
+transformers
+torch
+huggingface-hub

requirements.txt

@@ -3,6 +3,5 @@ tqdm
 requests>=2.20
 gradio==4.1.1
 openai>=1
+zhipuai
 sympy
-flask
-transformers