Hello-Happy-World

Hello-Happy-World is an AI-powered automatic dataset creation tool that supports generating LoRA datasets and SFT question sets from the web.

Features

• Automatic LoRA Dataset Generation: Generate LoRA datasets using pre-defined prompts.
• Automatic SFT Question Set Generation: Create SFT question sets from prompt files.
• DuckDuckGo Search Integration: Search specific topics via DuckDuckGo, and inject search results into the generated content.

Installation

Ensure you have installed the necessary libraries, such as requests and rich:

pip install requests rich

Usage

python main.py --task [LoRA/SFT] --output [output_file_name] --topic [search_topic]

• --task: Choose the type of generation, either LoRA or SFT.
• --output: Specify the output file name.
• --topic: Specify the topic for DuckDuckGo search. The search results will replace the {data} placeholder in the prompt file.

Examples

Generate an SFT question set, search for the topic "AI trends" on DuckDuckGo, and save it to sft_output.json:

python main.py --task SFT --output sft_output.json --topic "AI trends"

File Structure

• static/LoRA/prompt.md: Prompt file for generating LoRA datasets.
• static/SFT/prompt.md: Prompt file for generating SFT question sets.
• AI/config.yaml: Configuration file for the LLM model.

License

This project is licensed under the Apache License. For more details, see the LICENSE file.