soulteary commited on
Commit
77fbab7
1 Parent(s): a3cc1c9

docs: add demo

Browse files
Files changed (1) hide show
  1. README.md +14 -15
README.md CHANGED
@@ -14,7 +14,8 @@ widget:
14
  ---
15
  # Chinese Llama 2 7B
16
 
17
- 全部开源,完全可商用的**中文版 Llama2 模型及中英文 SFT 数据集**,输入格式严格遵循*llama-2-chat*的输入格式,以方便应用所有针对*llama-2-chat*的优化。
 
18
 
19
  ![Chinese LLaMA2 7B](.github/preview.jpg)
20
 
@@ -23,11 +24,18 @@ widget:
23
  > Talk is cheap, Show you the Demo.
24
 
25
  - [Demo 地址 / HuggingFace Spaces](https://huggingface.co/spaces/LinkSoul/Chinese-Llama-2-7b)
26
- -
27
- [Colab 一键启动](#
28
- )
 
 
 
 
 
 
29
 
30
  ## 快速测试
 
31
  ```python
32
  from transformers import AutoTokenizer, AutoModelForCausalLM, TextStreamer
33
 
@@ -41,18 +49,10 @@ instruction = """[INST] <<SYS>>\nYou are a helpful, respectful and honest assist
41
 
42
  If a question does not make any sense, or is not factually coherent, explain why instead of answering something not correct. If you don't know the answer to a question, please don't share false information.\n<</SYS>>\n\n{} [/INST]"""
43
 
44
- prompt = instruction.format("用中文回答,When is the best time to visit Beijing, and do you have any suggestions for me?")
45
  generate_ids = model.generate(tokenizer(prompt, return_tensors='pt').input_ids.cuda(), max_new_tokens=4096, streamer=streamer)
46
  ```
47
 
48
- ## 资源下载
49
-
50
- - 模型下载:[Chinese Llama2 Chat Model](https://huggingface.co/LinkSoul/Chinese-Llama-2-7b)
51
-
52
- > 我们使用了中英文 SFT 数据集,数据量 1000 万。
53
-
54
- - 数据集:[https://huggingface.co/datasets/LinkSoul/instruction_merge_set](https://huggingface.co/datasets/LinkSoul/instruction_merge_set)
55
-
56
  ## 如何训练
57
 
58
  ```bash
@@ -61,9 +61,8 @@ python train.py --args ...
61
 
62
  ## 相关项目
63
 
64
- - [Llama2](#)
65
 
66
  ## 项目协议
67
 
68
-
69
  [Apache-2.0 license](https://github.com/LinkSoul-AI/Chinese-Llama-2-7b/blob/main/LICENSE)
 
14
  ---
15
  # Chinese Llama 2 7B
16
 
17
+
18
+ 全部开源,完全可商用的**中文版 Llama2 模型及中英文 SFT 数据集**,输入格式严格遵循 *llama-2-chat* 格式,兼容适配所有针对原版 *llama-2-chat* 模型的优化。
19
 
20
  ![Chinese LLaMA2 7B](.github/preview.jpg)
21
 
 
24
  > Talk is cheap, Show you the Demo.
25
 
26
  - [Demo 地址 / HuggingFace Spaces](https://huggingface.co/spaces/LinkSoul/Chinese-Llama-2-7b)
27
+ - [Colab 一键启动](#) // 正在准备
28
+
29
+ ## 资源下载
30
+
31
+ - 模型下载:[Chinese Llama2 Chat Model](https://huggingface.co/LinkSoul/Chinese-Llama-2-7b)
32
+
33
+ > 我们使用了中英文 SFT 数据集,数据量 1000 万。
34
+
35
+ - 数据集:[https://huggingface.co/datasets/LinkSoul/instruction_merge_set](https://huggingface.co/datasets/LinkSoul/instruction_merge_set)
36
 
37
  ## 快速测试
38
+
39
  ```python
40
  from transformers import AutoTokenizer, AutoModelForCausalLM, TextStreamer
41
 
 
49
 
50
  If a question does not make any sense, or is not factually coherent, explain why instead of answering something not correct. If you don't know the answer to a question, please don't share false information.\n<</SYS>>\n\n{} [/INST]"""
51
 
52
+ prompt = instruction.format("用英文回答,什么是夫妻肺片?")
53
  generate_ids = model.generate(tokenizer(prompt, return_tensors='pt').input_ids.cuda(), max_new_tokens=4096, streamer=streamer)
54
  ```
55
 
 
 
 
 
 
 
 
 
56
  ## 如何训练
57
 
58
  ```bash
 
61
 
62
  ## 相关项目
63
 
64
+ - [Llama2](https://ai.meta.com/llama/)
65
 
66
  ## 项目协议
67
 
 
68
  [Apache-2.0 license](https://github.com/LinkSoul-AI/Chinese-Llama-2-7b/blob/main/LICENSE)