Skip to content

Commit

Permalink
update new training config and Tutorial
Browse files Browse the repository at this point in the history
  • Loading branch information
chg0901 committed Apr 29, 2024
1 parent 942429c commit e94028f
Showing 1 changed file with 24 additions and 4 deletions.
28 changes: 24 additions & 4 deletions xtuner_config/README_llama3_8b_instruct_qlora_alpaca_e3_M.md
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,27 @@
- 修改Xtuner模型配置文件
- 在EmoLLM项目上进行基于Xtuner进行QLoRA微调

## 更新

已经上传了最新的训练配置文件, 进行了些许改动, 训练数据中添加了85条自我认知数据和240条弱智吧数据.

### 更新的文件

- 配置文件[llama3_8b_instruct_qlora_alpaca_e3_M_ruozhi_scM](./llama3_8b_instruct_qlora_alpaca_e3_M_ruozhi_scM.py)
- [弱智吧原始数据**ruozhiba_raw.jsonl**](../datasets/ruozhiba_raw.jsonl)
- [弱智吧原始数据的Python处理文件**ruozhiba_raw_data_process.py**](../datasets/ruozhiba_raw_data_process.py)
- [ruozhiba_raw_data_process.py处理之后的弱智吧数据**ruozhiba_format_emo.jsonl**](../datasets/processed/ruozhiba_format_emo.jsonl)
- [数据集划分工具代码**split_dataset.py**](../datasets/split_dataset.py)
- [调用split_dataset.py的示例代码**split_shuffle.py**](../datasets/split_shuffle.py)

### 更新的有关参考教程

请参考以下知乎链接进行训练和测评

- [[Llama3][InternLM2][RuoZhiBa][EmoLLM]**使用弱智吧数据微调Llama3-Instruct-8B模型**](https://zhuanlan.zhihu.com/p/694818596)
- [[Llama3][EmoLLM][Minisora]**Meta Llama 3快速上手:用EmoLLM数据基于Xtuner采用QLoRA微调Meta-Llama-3-8B-Instruct模型**【V1】](https://zhuanlan.zhihu.com/p/693454096)
- [[Llama3][InternLM2]**OpenCompass 大模型评测Llama3-instruct-8B有关模型**](https://zhuanlan.zhihu.com/p/694922988)

## 模型和有关GitHub项目下载

### Llama-3-8B-Instruct模型下载
Expand Down Expand Up @@ -41,7 +62,7 @@ conda create -n llama python=3.10
conda install pytorch==2.1.0 torchvision==0.16.0 torchaudio==2.1.0 pytorch-cuda=12.1 -c pytorch -c nvidia
```

### git clone xtuner-0.1.18.dev0
### git clone XTuner-0.1.18

```python
git clone https://github.com/InternLM/xtuner
Expand All @@ -68,7 +89,7 @@ llama3_chat=dict(
3. 在双换行 \n\n 之后,消息的内容随之而来。每条消息的结尾由 <|eot_id|> 令牌标记。
- Ref: [ArtificialZeng/llama3_explained](https://github.com/ArtificialZeng/llama3_explained)

### 安装xtuner-0.1.18.dev0
### 安装XTuner-0.1.18

```python
# 进入源码目录
Expand Down Expand Up @@ -498,5 +519,4 @@ python cli_Llama3.py

### **知乎原文**

1. [Llama3][EmoLLM][Minisora]Meta Llama 3快速上手:用EmoLLM数据基于Xtuner采用QLoRA微调Meta-Llama-3-8B-Instruct模型【V0】 - 知乎 https://zhuanlan.zhihu.com/p/693321573
2. [Llama3][EmoLLM][Minisora]Meta Llama 3快速上手:用EmoLLM数据基于Xtuner采用QLoRA微调Meta-Llama-3-8B-Instruct模型【V1】 - 知乎 https://zhuanlan.zhihu.com/p/693454096
1. [Llama3][EmoLLM][Minisora]Meta Llama 3快速上手:用EmoLLM数据基于Xtuner采用QLoRA微调Meta-Llama-3-8B-Instruct模型【V1】 - 知乎 https://zhuanlan.zhihu.com/p/693454096

0 comments on commit e94028f

Please sign in to comment.