Skip to content

Commit

Permalink
Merge pull request LianjiaTech#359 from Life-0-1/main
Browse files Browse the repository at this point in the history
Update README.md -- add news: BELLE-LLaMA-EXT-13B
  • Loading branch information
Life-0-1 authored May 14, 2023
2 parents 2297f6b + 5d51471 commit 50baaab
Showing 1 changed file with 5 additions and 0 deletions.
5 changes: 5 additions & 0 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -22,6 +22,7 @@
</br>

## 🔄 最近更新
* [2023/05/14] 开放[BELLE-LLaMA-EXT-13B](https://huggingface.co/BelleGroup/BELLE-LLaMA-EXT-13B),在LLaMA-13B的基础上扩展中文词表,并在400万高质量的对话数据上进行训练。
* [2023/05/11] [BELLE/data/10M](data/10M)中,新加350万条生成多样化指令任务数据,包括单轮和多轮对话[train_3.5M_CN](https://huggingface.co/datasets/BelleGroup/train_3.5M_CN)
* [2023/04/19] 开放了其中一篇论文中的的相关模型:包括在LLaMA7B基础上增量预训练扩展中文词表的模(详见[BelleGroup/BELLE-LLaMA-EXT-7B](https://huggingface.co/BelleGroup/BELLE-LLaMA-EXT-7B)),以及基于多样化开源数据训练后的LLaMA-7B模型(详见[BelleGroup/BELLE-on-Open-Datasets](https://huggingface.co/BelleGroup/BELLE-on-Open-Datasets))。
* [2023/04/18] 更新了train代码,详见[BELLE/train](https://github.com/LianjiaTech/BELLE/tree/main/train),集成了Deepspeed-Chat,提供了相关的docker
Expand Down Expand Up @@ -279,3 +280,7 @@ LLaMA-7B + FT(2M) + LoRA(math_0.25M) 代表了一个在0.25M数学指令数据
## ☎️ 联系我们

欢迎大家来[Discord](https://discord.gg/pMPY53UUGq)[微信](https://github.com/LianjiaTech/BELLE/blob/main/assets/belle_wechat.jpg)与我们交流。

## ⭐️ Star History

[![Star History Chart](https://api.star-history.com/svg?repos=LianjiaTech/BELLE&type=Date)](https://star-history.com/#LianjiaTech/BELLE&Date)

0 comments on commit 50baaab

Please sign in to comment.