LLM Kit Download - LLM Kit Source code download

LLM Kit

AI Source Code

1.0.0

Download

Suqin (suqin) Language large model full tool chain WebUI integration package

English |中文

introduce

⛓️ The goal of this project is to implement a full-process tool WebUI integration package for all major language models. You can have your own customized models and exclusive applications without writing code!

The specific function modules are as follows:

Development and deployment

For first-time users, it is recommended to consult the documentation for related operations.

Software requirements

This project has been tested in Python 3.8 - 3.10, CUDA 11.7/11.8 environment. Testing has been completed on Windows and Linux systems.

Installation environment

1. Operating environment

 # 拉取仓库
$ git clone https://github.com/wpydcr/LLM-Kit.git

# 进入目录
$ cd LLM-Kit

# 安装依赖
$ pip install -r requirements.txt

Dependencies can also be downloaded directly: Windows environment package, Linux environment package.

Install the graphics card driver and cuda yourself

2. Execute the script

In win environment, double-click to open the web-demo-CN.bat script
Linux environment, double-click to open the web-demo-CN.sh script

3. Related presentation files

To experience the large model connection database function, please download the relevant files first.
To experience the role-playing function, please download the relevant files first
For the expression pack function, please download the relevant files first and put them in the data/apply/emotion folder.

File directory

env : integrated package environment
utils : tool code
modules : each module code
- agent : agent related code
  - chatdb : database access related code
  - svc : svc related code
  - vits : vits related codes
- apply : apply demo code
- model : model related code
data : data file
- apply : apply demo related files
  - audio : generated voice file
  - emotion : expression package file
  - play : character prompt word file
- documents : knowledge base data files
- modeldata : model training data file
  - LLM : Large model training data file
  - Embedding : Embed model training data file
ui : ui interface code
output : checkpoints in training
models : model files
- LLM : Language large model file
- Embedding : Embedding model files
- LoRA : LoRA model file
- live2d_models : live2d model files
- svc_models : svc related files
  - hubert_model : sound encoder model file
  - svc : svc model file
- vits_pretrained_models : vits model file

Development roadmap

Contributor

Wu Pingyu@juncaisuyun

Initiator, responsible for the overall project architecture and technical planning

Liao Wenbin@Institute of Computing Technology, Chinese Academy of Sciences

Responsible for python development such as gradient, graph vector database, Mysql database, API interface integration, etc.

Hu Jucheng@University College LondonUCL

Responsible for graph vector database, live2D, vits, svc, gradient and other python development

Li Jinghui

Responsible for back-end code development of large model training and inference

Project recruitment

See documentation for details

Reference and Learning

Langchain application based on ChatGLM

Camel Chinese model

ChatDB

Live2d model

Chat-Haruhi-Suzumiya

Chuanhu Chat: Provides web graphical interface for various LLMs

Open source agreement

The code of this repository is open source under the AGPL-3.0 agreement

On the one hand, we hope to strengthen the productization of the project through open source; on the other hand, we hope to absorb more practical scenarios in the community and continue to improve the product. We also welcome everyone to participate in the project.

AGPL-3.0 is an OSI-approved license that complies with all standards of free and open source software. Open source will always be our original intention and core. We will always persist in doing this. We believe that with the promotion of the community, this We will definitely do something better.

Perhaps many developers have some questions about this protocol. The open source community currently has many open source software that adopt the AGPL-3.0 protocol, such as MongoDB, Grafana, Loki, etc. Wikipedia also has a list of which open source projects have adopted AGPL. -3.0 open source license.

The AGPL-3.0 agreement has a very critical point, that is, the secondary distribution version after modifying the code of the upstream open source project must also be open source. The agreement restricts some companies who want to Folk open source project code for closed source commercial distribution, and follow the upstream open source The project's maintenance team engages in direct commercial competition. If it is only used within the enterprise without modification at any level, users do not need to worry about the restrictions brought by the AGPL-3.0 agreement. These conditions are designed to encourage third parties who want to modify the software. Also contribute to the project and community. We believe this is a more equitable way forward, and we believe it will help us build stronger communities.

To put it simply: if you modify the source code of this project, you must contribute these modifications to the community. Modified and derived codes are never allowed to be released and sold as closed source commercial software.

We also provide commercial authorization. If you need to re-develop, change and use this product for commercial purposes, please contact us ([email protected]) for commercial authorization to comply with AGPL-3.0 The agreement guarantees your normal use.

In addition to this, we will also accept customized development based on personal or corporate needs at our discretion.

At present, the domestic GPL agreement has the characteristics of a contract and is a civil legal act, which falls within the scope of my country's Contract Law. The project team reserves the right to litigate.

The open source team of this project has the final right to interpret this open source agreement.

Quote

If you use our models, code or data in your projects, please cite the following article.

Please cite the repo if you use the data or code in this repo.

 @misc{wupingyu2023,
  author={Pingyu Wu},
  title = {LLM Kit},
  year = {2023},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {url{https://github.com/wpydcr/LLM-Kit.git}},
}