Baicai003 commited on
Commit
b730de1
1 Parent(s): 63250c7

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +13 -9
README.md CHANGED
@@ -12,20 +12,24 @@ tags:
12
  - llama3
13
  - chat
14
  ---
15
- # llama3-instruct 中文DPO版
16
  ### 模型介绍
17
  像原版instruct一样,喜欢用有趣中文和表情符号回答问题。
18
  Github:https://github.com/CrazyBoyM/llama3-Chinese-chat
19
- 放出训练配方细节供网友参考分享:
20
- DPO(beta 0.5) + lora rank128, alpha256 + 打开"lm_head", "input_layernorm", "post_attention_layernorm", "norm"层训练.
21
- 特点:偏好中文和emoji表情,且不损伤原instruct版模型能力。实测中文DPO版问答性能体验超过现在市面上任何llama3中文微调版 (微调会破坏llama3原版能力,导致遗忘)
22
  ![Alt text](https://modelscope.cn/api/v1/models/baicai003/Llama3-Chinese-instruct-DPO-beta0.5/repo?Revision=master&FilePath=image.png&View=true)
23
 
 
 
 
24
  ### 模型部署
25
- 网页脚本文件:https://github.com/CrazyBoyM/llama3-Chinese-chat/blob/main/deploy/web_streamlit_for_instruct_v2.py
 
26
  ```
27
  pip install streamlit
28
- streamlit run web_streamlit_for_instruct_v2.py ./Llama3-Chinese-instruct-DPO-beta0.5
 
29
  ```
30
 
31
  ### 模型下载
@@ -37,10 +41,10 @@ pip install modelscope
37
  ```python
38
  #SDK模型下载
39
  from modelscope import snapshot_download
40
- model_dir = snapshot_download('baicai003/Llama3-Chinese-instruct-DPO-beta0.5')
41
  ```
42
  Git下载
43
  ```
44
  #Git模型下载
45
- git clone https://www.modelscope.cn/baicai003/Llama3-Chinese-instruct-DPO-beta0.5.git
46
- ```
 
12
  - llama3
13
  - chat
14
  ---
15
+ # llama3.1-8b-instruct 中文DPO版
16
  ### 模型介绍
17
  像原版instruct一样,喜欢用有趣中文和表情符号回答问题。
18
  Github:https://github.com/CrazyBoyM/llama3-Chinese-chat
19
+
20
+ 特点:偏好中文和emoji表情,且不损伤原instruct版模型能力。实测中文DPO版问答性能体验超过现在市面上任何llama3.1中文微调版 (微调会大面积破坏llama3.1原版能力,导致遗忘)
 
21
  ![Alt text](https://modelscope.cn/api/v1/models/baicai003/Llama3-Chinese-instruct-DPO-beta0.5/repo?Revision=master&FilePath=image.png&View=true)
22
 
23
+ ### 训练细节
24
+ DPO(beta 0.5) + lora rank128, alpha256 + 打开"lm_head", "input_layernorm", "post_attention_layernorm", "norm"层训练.
25
+
26
  ### 模型部署
27
+ 网页脚本文件:https://github.com/CrazyBoyM/llama3-Chinese-chat/blob/main/deploy/web_streamlit_for_instruct_v2.py
28
+ 已经在模型文件夹里内置了一份web.py,可以直接使用。
29
  ```
30
  pip install streamlit
31
+ pip install transformers==4.40.1
32
+ streamlit run web.py ./llama3.1-8b-instruct-dpo-zh
33
  ```
34
 
35
  ### 模型下载
 
41
  ```python
42
  #SDK模型下载
43
  from modelscope import snapshot_download
44
+ model_dir = snapshot_download('shareAI/llama3.1-8b-instruct-dpo-zh')
45
  ```
46
  Git下载
47
  ```
48
  #Git模型下载
49
+ git clone https://www.modelscope.cn/shareAI/llama3.1-8b-instruct-dpo-zh.git
50
+ ```