---
license: apache-2.0
---
### Instruction
this model is trained on 50G chinese and english instruction datasets,the Moses25/Llama-3-8B-chat-32K is based on it with about 180K instruction dataset. 
You can download Moses25/Instruct-dataset11M dataset to train your custom model.The github repo is [fintune-repo](https://github.com/moseshu/llama-recipes)

```
git clone https://github.com/moseshu/llama-recipes
cd llama-recipes/src && sh ft.sh
```