https://github.com/qew21/autoleetcode

Automatically read Leetcode problems using Selenium, hand them over to different large models for processing, and automatically submit solutions while tracking accuracy rates.
https://github.com/qew21/autoleetcode

leetcode llm qwen2 selenium

Last synced: 12 days ago
JSON representation

Automatically read Leetcode problems using Selenium, hand them over to different large models for processing, and automatically submit solutions while tracking accuracy rates.

• Host: GitHub
• URL: https://github.com/qew21/autoleetcode
• Owner: qew21
• Created: 2024-06-13T15:41:55.000Z (7 months ago)
• Default Branch: master
• Last Pushed: 2024-07-03T15:28:05.000Z (6 months ago)
• Last Synced: 2024-11-06T20:15:45.894Z (2 months ago)
• Topics: leetcode, llm, qwen2, selenium
• Language: Python
• Homepage:
• Size: 178 KB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Autoleetcode

#### 介绍

使用selenium自动读取Leetcode题目，交给不同的大模型处理，自动提交统计正确率

目前看顶尖的大模型尚不能高正确率的通过中等难度题目，预期通过微调AlchemistCoder提高准确率

#### 使用说明

1. 下载匹配Chrome浏览器版本的[driver](https://googlechromelabs.github.io/chrome-for-testing/#stable)

2. 将driver放到当前目录下

3. 参考如下配置修改配置中的ApiKey和BaseUrl等参数

4. 安装依赖，运行 ```python main.py```

5. 处于登录状态时，会自动跳转到登录页面，登录成功后会自动跳转到题目页面

```json5

{

    "BaseUrl": "https://dashscope.aliyuncs.com/compatible-mode/v1",  // dashscope api url

    "ApiKey": "sk-****",  // dashscope api key

    "Model": "qwen2-72b-instruct",  // model name

    "Level": "hard",  // difficulty level

    "Skip": false,  // skip already solved problems

    "Count": 100,  // problem count

    "StartPage": 1  // start page

}

```

#### 示例

![demo](demo.jpg)

---

#### Introduction

Automatically read Leetcode problems using Selenium, hand them over to different large models for processing, and automatically submit solutions while tracking accuracy rates.

It appears that even top-tier large models struggle to achieve high accuracy on medium difficulty problems. The expectation is to improve accuracy by fine-tuning the AlchemistCoder model.

#### Instructions

1. Download the driver [here](https://googlechromelabs.github.io/chrome-for-testing/#stable) that matches your Chrome browser version.

2. Place the driver in the current directory.

3. Create a `config.json` file based on example below, modifying parameters such as ApiKey and BaseUrl.

4. Install requirements and run ```python main.py```

5. When in a logged-out state, the system will automatically redirect to the login page; after successful login, it will automatically navigate to the problem page.