README.md

---
license: other
---
# LLaMA2-Accessory: An Open-source Toolkit for LLM Development 🚀

🚀**LLaMA2-Accessory** is an open-source toolkit for pre-training, fine-tuning and deployment of **Large Language Models (LLMs)** and **mutlimodal LLMs**. This repo is mainly inherited from [LLaMA-Adapter](https://github.com/OpenGVLab/LLaMA-Adapter) with more advanced features.🧠

<p align="left">
   Github link: <a href="https://github.com/Alpha-VLLM/LLaMA2-Accessory" target="_blank">Github</a> • 👋 join our <a href="https://github.com/Alpha-VLLM/LLaMA2-Accessory/blob/main/docs/wechat.md" target="_blank">WeChat</a>
</p>


## Features
* **💡Support More Datasets and Tasks**

  - 🎯 Pre-training with [RefinedWeb](https://huggingface.co/datasets/tiiuae/falcon-refinedweb) and [StarCoder](https://github.com/bigcode-project/starcoder).
  - 📚 Single-modal fine-tuning with [Alpaca](https://github.com/tatsu-lab/stanford_alpaca), [ShareGPT](https://github.com/domeccleston/sharegpt), [LIMA](https://arxiv.org/pdf/2305.11206.pdf), [UltraChat](https://github.com/thunlp/UltraChat) and [MOSS](https://github.com/OpenLMLab/MOSS).
  - 🌈 Multi-modal fine-tuning with image-text pairs ([LAION](https://laion.ai/blog/laion-5b/), [COYO](https://github.com/kakaobrain/coyo-dataset) and more), interleaved image-text data ([MMC4](https://github.com/allenai/mmc4) and [OBELISC](https://github.com/huggingface/OBELISC)) and visual instruction data ([LLaVA](https://github.com/haotian-liu/LLaVA), [Shrika](https://github.com/shikras/shikra), [Bard](https://bard.google.com/))
  - 🔧 LLM for API Control ([GPT4Tools](https://github.com/StevenGrove/GPT4Tools) and [Gorilla](https://github.com/ShishirPatil/gorilla)).
* **⚡Efficient Optimization and Deployment**

  - 🚝 Parameter-efficient fine-tuning with [Zero-init Attenion](https://github.com/OpenGVLab/LLaMA-Adapter) and [Bias-norm Tuning](https://github.com/OpenGVLab/LLaMA-Adapter).
  - 💻 Fully Sharded Data Parallel ([FSDP](https://engineering.fb.com/2021/07/15/open-source/fsdp/)), [Flash Attention 2](https://github.com/Dao-AILab/flash-attention) and [QLoRA](https://github.com/artidoro/qlora).
* **🏋️‍♀️Support More Visual Encoders and LLMs**

  - 👁‍🗨 Visual Encoders: [CLIP](https://github.com/openai/CLIP), [Q-Former](https://github.com/salesforce/LAVIS) and [ImageBind](https://github.com/facebookresearch/ImageBind).
  - 🧩 LLMs: LLaMA and LLaMA2.


## Installation

See [docs/install.md](./docs/install.md). 

## Training & Inference
See [docs/pretrain.md](./docs/pretrain.md) and [docs/finetune.md](./docs/finetune.md). 

## Demos
* Instruction-tuned LLaMA2: [alpaca](https://alpha-vllm.github.io/demo_presentation/examples/finetune/sg/alpaca.html) & [gorilla](https://alpha-vllm.github.io/demo_presentation/examples/finetune/sg/gorilla.html).
* Chatbot LLaMA2: [dialog_sharegpt](https://alpha-vllm.github.io/demo_presentation/examples/finetune/sg/dialog_sharegpt.html) & [dialog_lima](https://alpha-vllm.github.io/demo_presentation/examples/finetune/sg/dialog_lima.html) & [llama2-chat](https://alpha-vllm.github.io/demo_presentation/examples/finetune/sg/llama2-chat.html).
* Multimodal LLaMA2: [in-context](https://alpha-vllm.github.io/demo_presentation/examples/finetune/mm/in-context.html)


## Core Contributors

[Chris Liu](https://github.com/ChrisLiu6), [Ziyi Lin](https://github.com/linziyi96), [Guian Fang](https://github.com/Enderfga), [Jiaming Han](https://github.com/csuhan), [Renrui Zhang](https://github.com/ZrrSkywalker), [Wenqi Shao](https://github.com/wqshao126), [Peng Gao](https://github.com/gaopengpjlab)

## Hiring Announcement

🔥 **We are hiring** interns, postdocs, and full-time researchers at the **General Vision Group, Shanghai AI Lab**, with a focus on multi-modality and vision foundation models. If you are interested, please contact [gaopengcuhk@gmail.com](mailto:gaopengcuhk@gmail.com).

## Citation
If you find our code and paper useful, please kindly cite:
```bash
@article{zhang2023llamaadapter,
  title = {LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention},
  author={Zhang, Renrui and Han, Jiaming and Liu, Chris and Gao, Peng and Zhou, Aojun and Hu, Xiangfei and Yan, Shilin and Lu, Pan and Li, Hongsheng and Qiao, Yu},
  journal={arXiv preprint arXiv:2303.16199},
  year={2023}
}
```
```bash
@article{gao2023llamaadapterv2,
  title = {LLaMA-Adapter V2: Parameter-Efficient Visual Instruction Model},
  author={Gao, Peng and Han, Jiaming and Zhang, Renrui and Lin, Ziyi and Geng, Shijie and Zhou, Aojun and Zhang, Wei and Lu, Pan and He, Conghui and Yue, Xiangyu and Li, Hongsheng and Qiao, Yu},
  journal={arXiv preprint arXiv:2304.15010},
  year={2023}
}
```

## Acknowledgement
+ [@facebookresearch](https://github.com/facebookresearch) for [llama](https://github.com/facebookresearch/llama)
+ [@OpenGVLab](https://github.com/OpenGVLab) for [LLaMA-Adapter](https://github.com/OpenGVLab/LLaMA-Adapter)
+ [@facebookresearch](https://github.com/facebookresearch) for [ImageBind](https://github.com/facebookresearch/ImageBind) & [LIMA](https://huggingface.co/datasets/64bits/lima_vicuna_format)
+ [@Instruction-Tuning-with-GPT-4](https://github.com/Instruction-Tuning-with-GPT-4) for [GPT-4-LLM](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM)
+ [@tatsu-lab](https://github.com/tatsu-lab) for [stanford_alpaca](https://github.com/tatsu-lab/stanford_alpaca)
+ [@tloen](https://github.com/tloen) for [alpaca-lora](https://github.com/tloen/alpaca-lora)
+ [@lm-sys](https://github.com/lm-sys) for [FastChat](https://github.com/lm-sys/FastChat)
+ [@domeccleston](https://github.com/domeccleston) for [sharegpt](https://github.com/domeccleston/sharegpt)
+ [@karpathy](https://github.com/karpathy) for [nanoGPT](https://github.com/karpathy/nanoGPT)
+ [@Dao-AILab](https://github.com/Dao-AILab) for [flash-attention](https://github.com/Dao-AILab/flash-attention)
+ [@NVIDIA](https://github.com/NVIDIA) for [apex](https://github.com/NVIDIA/apex) & [Megatron-LM](https://github.com/NVIDIA/Megatron-LM)
+ [@Vision-CAIR](https://github.com/Vision-CAIR) for [MiniGPT-4](https://github.com/Vision-CAIR/MiniGPT-4)
+ [@haotian-liu](https://github.com/haotian-liu) for [LLaVA](https://github.com/haotian-liu/LLaVA)
+ [@huggingface](https://github.com/huggingface) for [peft](https://github.com/huggingface/peft) & [OBELISC](https://github.com/huggingface/OBELISC)
+ [@Lightning-AI](https://github.com/Lightning-AI) for [lit-gpt](https://github.com/Lightning-AI/lit-gpt) & [lit-llama](https://github.com/Lightning-AI/lit-llama)
+ [@allenai](https://github.com/allenai) for [mmc4](https://github.com/allenai/mmc4)
+ [@StevenGrove](https://github.com/StevenGrove) for [GPT4Tools](https://github.com/StevenGrove/GPT4Tools)
+ [@ShishirPatil](https://github.com/ShishirPatil) for [gorilla](https://github.com/ShishirPatil/gorilla)
+ [@OpenLMLab](https://github.com/OpenLMLab) for [MOSS](https://github.com/OpenLMLab/MOSS)
+ [@thunlp](https://github.com/thunlp) for [UltraChat](https://github.com/thunlp/UltraChat)
+ [@LAION-AI](https://github.com/LAION-AI) for [LAION-5B](https://laion.ai/blog/laion-5b/)
+ [@shikras](https://github.com/shikras) for [shikra](https://github.com/shikras/shikra)
+ [@kakaobrain](https://github.com/kakaobrain) for [coyo-dataset](https://github.com/kakaobrain/coyo-dataset)
+ [@salesforce](https://github.com/salesforce) for [LAVIS](https://github.com/salesforce/LAVIS)
+ [@openai](https://github.com/openai) for [CLIP](https://github.com/openai/CLIP)
+ [@bigcode-project](https://github.com/bigcode-project) for [starcoder](https://github.com/bigcode-project/starcoder)
+ [@tiiuae](https://huggingface.co/tiiuae) for [falcon-refinedweb](https://huggingface.co/datasets/tiiuae/falcon-refinedweb)
+ [@microsoft](https://github.com/microsoft) for [DeepSpeed](https://github.com/microsoft/DeepSpeed)
+ [@declare-lab](https://github.com/declare-lab) for [flacuna](https://github.com/declare-lab/flacuna)
+ [@Google](https://github.com/google) for [Bard](https://bard.google.com/)

## License

Llama 2 is licensed under the [LLAMA 2 Community License](LICENSE_llama2), Copyright (c) Meta Platforms, Inc. All Rights Reserved.