Easy Dataset × LLaMA Factory: Enabling LLMs to Efficiently Learn Domain Knowledge

输入“/”快速插入内容

📚
Easy Dataset × LLaMA Factory: Enabling LLMs to Efficiently Learn Domain Knowledge

2025年5月20日修改

5564

7024

Easy Dataset is an application designed for creating fine-tuning datasets for large language models (LLMs). It provides an intuitive interface for uploading domain-specific files, intelligently segmenting content, generating questions, and producing high-quality training data for model fine-tuning. It supports LLM APIs such as OpenAI, DeepSeek, Volcengine, and Ollama local APIs.

LLaMA Factory is an open-source low-code fine-tuning framework for large models, which integrates the most widely used fine-tuning techniques in the industry. It supports fine-tuning large models without coding through a Web UI interface. Currently, it has become one of the most popular fine-tuning frameworks in the open-source community, with more than 46,000 stars on GitHub. It supports full fine-tuning, LoRA fine-tuning, as well as fine-tuning algorithms such as SFT and DPO.

This tutorial uses Easy Dataset to build SFT fine-tuning data from the public financial reports of five Internet companies, and uses LLaMA Factory to fine-tune the Qwen2.5-3B-Instruct model, so that the fine-tuned model can learn the knowledge in the financial report dataset.​

Running environment requirements

•
GPU Memory: Greater than or equal to 12 GB​

•
CUDA version: Higher than 11.6​

•
Python Version: 3.10​

Generate fine-tuning data using Easy Dataset

Install Easy Dataset

Method 1: Use the installation package

If your operating system is Windows, Mac, or a Unix system based on the ARM architecture, you can directly go to the Easy Dataset repository to download the installation package: https://github.com/ConardLi/easy-dataset/releases/latest

Method 2: Use Dockerfile

1.
Fetch the Easy Dataset repository from GitHub​

代码块

git clone https://github.com/ConardLi/easy-dataset.git​
cd easy-dataset​

2.
Build Docker mirroring​

代码块

docker build -t easy-dataset .

3.
Run a container​

代码块

docker run -d \​
    -p 1717:1717 \​
    -v {YOUR_LOCAL_DB_PATH}:/app/local-db \​
    --name easy-dataset \​
    easy-dataset​

Method 3: Install using NPM

1.
Download Node.js and pnpm​

Go to the Node.js and pnpm official websites to install the environment: https://nodejs.org/en/download | https://pnpm.io/

Use the following code to check if the Node.js version is higher than 18.0

代码块

node -v # v22.14.0

2.
Fetch the Easy Dataset repository from GitHub​

代码块

git clone https://github.com/ConardLi/easy-dataset.git​
cd easy-dataset​

3.
Install software dependencies​

代码块

pnpm install

4.
Launch the Easy Dataset application​

代码块

pnpm build​
pnpm start​

If the following output appears in the console, it indicates that the startup was successful. Open a browser and visit the corresponding website , and you can see the interface of Easy Dataset.

代码块

> easy-dataset@1.2.3 start​
> next start -p 1717​
​
  ▲ Next.js 14.2.25​
  - Local:        http://localhost:1717​
​
 ✓ Ready in 287ms​

Download sample data

This tutorial has prepared a batch of financial reports of Internet companies as sample data, including five reports of domestic Internet companies for the second quarter of 2024, in formats including txt and markdown. You can use git commands or directly visit the repository link to download.

代码块

git clone https://github.com/llm-factory/FinancialData-SecondQuarter-2024.git

All the data is in plain text, and the following is an example of the excerpted content.​

🥖

 Kuaishou's net profit increased by more than  than 70% in the second quarter, and CEO Cheng Yixiao emphasized the monetization of Kuaishou's AI technology​