KHAOSZ

English • 中文

A lightweight Transformer training & inference framework

📖 Table of Contents | 目录

English

• Installation
• Quick Start
• Documentation

中文

• 安装
• 快速开始
• 文档

---

English

Features

• 🚀 High Performance: Optimized for both training and inference
• 🔧 Flexible: Support for seq/sft/dpo training
• 💡 Easy to Use: Simple API with comprehensive examples
• 📦 Lightweight: Minimal dependencies

Installation

git clone https://github.com/username/khaosz.git
cd khaosz
pip install -e .

Quick Start

# Train
python tools/train.py \
  --train_type=seq \
  --data_root_path=/path/to/dataset \
  --param_path=/path/to/param_path

# Generate
python tools/generate.py --param_path=/path/to/param_path

Demo

# run download before using
python demo/download.py

# run demo
python demo/stream_chat.py
python demo/generate_batch.py
python demo/generate_ar.py

• bilibili

---

中文

特性

• 🚀 高性能: 训练与推理双向优化
• 🔧 灵活: 支持 seq/sft/dpo 多种训练方式
• 💡 易用: 简洁的 API 与丰富的示例
• 📦 轻量: 依赖少，部署简单

安装

git clone https://github.com/username/khaosz.git
cd khaosz
pip install -e .

快速开始

# 训练
python tools/train.py \
  --train_type=seq \
  --data_root_path=/path/to/dataset \
  --param_path=/path/to/param_path

# 生成
python tools/generate.py --param_path=/path/to/param_path

演示

# 使用前先下载模型
python demo/download.py

# 运行示例
python demo/stream_chat.py
python demo/generate_batch.py
python demo/generate_ar.py

• bilibili

---

Document | 文档

Document	Description
Parameter Guide 参数说明	Training & inference parameters 训练与推理参数配置
Design Document 设计文档	Framework architecture & module design 系统架构与模块设计
Data Flow 数据流程	Data processing pipeline details 数据处理管道详解
Model Introduction 模型介绍	Model architecture & technical details 模型架构与技术细节

Download | 下载

• HuggingFace
• python demo/download.py

Lincence | 许可证

• GPL-3.0