Update README.md
Browse files
README.md
CHANGED
@@ -1,5 +1,75 @@
|
|
1 |
-
---
|
2 |
-
license: other
|
3 |
-
license_name: license
|
4 |
-
license_link: LICENSE
|
5 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: other
|
3 |
+
license_name: license
|
4 |
+
license_link: LICENSE
|
5 |
+
---
|
6 |
+
<div align="center">
|
7 |
+
<h1>
|
8 |
+
Index-1.9B-Chat-GGUF
|
9 |
+
</h1>
|
10 |
+
</div>
|
11 |
+
|
12 |
+
|
13 |
+
此为[Index-1.9B-Chat](https://huggingface.co/IndexTeam/Index-1.9B-Chat)的GGUF版本,适配llama.cpp,同时提供了Ollma的ModelFile适配。
|
14 |
+
|
15 |
+
|
16 |
+
更多细节详见我们的[GitHub](https://github.com/bilibili/Index-1.9B)和[Index-1.9B技术报告](https://github.com/bilibili/Index-1.9B/blob/main/Index-1.9B%20%E6%8A%80%E6%9C%AF%E6%8A%A5%E5%91%8A.pdf)
|
17 |
+
|
18 |
+
### LLAMA.CPP
|
19 |
+
```shell
|
20 |
+
# 安装llama.cpp(https://github.com/ggerganov/llama.cpp)
|
21 |
+
git clone https://github.com/ggerganov/llama.cpp
|
22 |
+
cd llama.cpp
|
23 |
+
make
|
24 |
+
|
25 |
+
# 安装llama-cpp-python(https://github.com/abetlen/llama-cpp-python)
|
26 |
+
pip install llama-cpp-python
|
27 |
+
```
|
28 |
+
llama.cpp终端交互
|
29 |
+
```shell
|
30 |
+
./build/bin/llama-cli -m models/Index-1.9B-Chat/ggml-model-bf16.gguf --color -if
|
31 |
+
```
|
32 |
+
**注意!!** llama.cpp并不支持自定义chat_template, 故而需要自己拼接prompt,Index-1.9B的chat_template(示意)为
|
33 |
+
```shell
|
34 |
+
# 三个分隔符为 <unk>, reserved_0, reserved_1
|
35 |
+
[<unk>]sytem_message[reserved_0]user_message[reserved_1]response
|
36 |
+
```
|
37 |
+
使用llama-cpp-python可支持自定义的chat_template(已写入GGUF,可直接使用)
|
38 |
+
```python
|
39 |
+
from llama_cpp import Llama
|
40 |
+
|
41 |
+
model_path = "Index-1.9B-Chat-GGUF/ggml-model-Q6_K.gguf"
|
42 |
+
llm = Llama(model_path =model_path, verbose=True)
|
43 |
+
output = llm.create_chat_completion(
|
44 |
+
messages = [
|
45 |
+
{"role": "system", "content": "你是由哔哩哔哩自主研发的大语言模型,名为“Index”。你能够根据用户传入的信息,帮助用户完成指定的任务,并生成恰当的、符合要求的回复。"},
|
46 |
+
#{"role": "system", "content": "你需要扮演B站评论区老哥,用评论区阴阳怪气的话术回复,不要说你是AI"},
|
47 |
+
{"role": "user","content": "篮球和鸡有什么关系"}
|
48 |
+
]
|
49 |
+
)
|
50 |
+
print(output)
|
51 |
+
```
|
52 |
+
### OLLAMA
|
53 |
+
- 安装[Ollama](https://github.com/ollama/ollama)
|
54 |
+
```shell
|
55 |
+
curl -fsSL https://ollama.com/install.sh | sh
|
56 |
+
```
|
57 |
+
```shell
|
58 |
+
# 启动server
|
59 |
+
ollama serve
|
60 |
+
|
61 |
+
# 适配模型, OllamaModelFile中可修改模型文件和System Message
|
62 |
+
ollama create Index-1.9B-Chat -f Index-1.9B-Chat-GGUF/OllamaModelFile
|
63 |
+
|
64 |
+
# 启动终端
|
65 |
+
ollama run Index-1.9B-Chat
|
66 |
+
|
67 |
+
# web调用,可以动态指定System Message
|
68 |
+
curl http://localhost:11434/api/chat -d '{
|
69 |
+
"model": "Index-1.9B-Chat",
|
70 |
+
"messages": [
|
71 |
+
{ "role": "system", "content": "你是由哔哩哔哩自主研发的大语言模型,名为“Index”。你能够根据用户传入的信息,帮助用户完成指定的任务,并生成恰当的、符合要求的回复。" },
|
72 |
+
{ "role": "user", "content": "续写 金坷垃" }
|
73 |
+
]
|
74 |
+
}'
|
75 |
+
```
|