README.md 4.08 KB
Newer Older
chenych's avatar
chenych committed
1
# 模型名称(跟原生模型一致)
chenzk's avatar
v1.0  
chenzk committed
2
## 论文
chenych's avatar
chenych committed
3
4
`此处填写论文名称`
- 此处填写论文地址链接
liuhy's avatar
liuhy committed
5

chenych's avatar
chenych committed
6
如果没有写`暂无`
chenzk's avatar
v1.0.6  
chenzk committed
7

chenych's avatar
chenych committed
8
9
## 模型简介
简要介绍模型结构,根据论文或者原生模型介绍内容填写,如果有模型结构或者模型算法图则放图,没有则不放。
liuhy's avatar
liuhy committed
10

chenzk's avatar
v1.0.6  
chenzk committed
11
12
13
14
<div align=center>
    <img src="./doc/xxx.png"/>
</div>

chenych's avatar
chenych committed
15
16
17
18
19
20
21
22
23
24
25
26
27
## 环境依赖
- 列举基础环境需求,根据实际情况填写

| 软件 | 版本 |
| :------: | :------: |
| DTK | xxx |
| python | xx |
| transformers | xx |
| vllm | xx |
| paddlepaddle | xx |

推荐使用镜像:
- `docker_name``imageID`根据实际模型情况修改
liuhy's avatar
liuhy committed
28

chenzk's avatar
v1.0  
chenzk committed
29
```
chenych's avatar
chenych committed
30
docker run -it --shm-size 200g --network=host --name `docker_name` --privileged --device=/dev/kfd --device=/dev/dri --device=/dev/mkfd --group-add video --cap-add=SYS_PTRACE --security-opt seccomp=unconfined -u root -v /path/your_code_data/:/path/your_code_data/ -v /opt/hyhal/:/opt/hyhal/:ro `imageID` bash
chenzk's avatar
v1.0  
chenzk committed
31
```
chenych's avatar
chenych committed
32
更多镜像可前往[光源](https://sourcefind.cn/#/service-list)下载使用。
liuhy's avatar
liuhy committed
33

chenych's avatar
chenych committed
34
关于本项目DCU显卡所需的特殊深度学习库可从[光合](https://developer.sourcefind.cn/tool/)开发者社区下载安装,其它包参照requirements.txt安装:
chenzk's avatar
v1.0  
chenzk committed
35
36
37
```
pip install -r requirements.txt
```
chenzk's avatar
chenzk committed
38

chenych's avatar
chenych committed
39
40
## 数据集
[公开数据集名称](公开数据集官网下载地址,过小文件可打包到项目里。)
liuhy's avatar
liuhy committed
41

chenzk's avatar
v1.0  
chenzk committed
42
43
44
45
此处提供数据预处理脚本的使用方法
```
python xxx.py
```
chenzk's avatar
v1.0.1  
chenzk committed
46
项目中已提供用于试验训练的迷你数据集,训练数据目录结构如下,用于正常训练的完整数据集请按此目录结构进行制备:
chenzk's avatar
v1.0  
chenzk committed
47
48
49
50
51
52
53
54
55
56
57
```
 ── dataset
    │   ├── label_1
    │             ├── xxx.png
    │             ├── xxx.png
    │             └── ...
    │   └── label_2
    │             ├── xxx.png
    │             ├── xxx.png
    │             └── ...
```
chenych's avatar
chenych committed
58
59
如果没有数据集,写`暂无`

chenzk's avatar
v1.0  
chenzk committed
60
## 训练
chenych's avatar
chenych committed
61
62
63
`单机训练``多机训练`方法根据实际情况选择填写即可。
如果没有训练脚本,则写`暂无`,后面`单机训练``多机训练`章节删掉。
### 单机训练
chenzk's avatar
v1.0  
chenzk committed
64
```
chenzk's avatar
v1.0.1  
chenzk committed
65
sh xxx.sh # 或python xxx.py
chenzk's avatar
v1.0  
chenzk committed
66
```
liuhy's avatar
liuhy committed
67

chenych's avatar
chenych committed
68
### 多机训练
chenzk's avatar
v1.0  
chenzk committed
69
```
chenzk's avatar
v1.0.1  
chenzk committed
70
sh xxx.sh 或python xxx.py
chenzk's avatar
v1.0  
chenzk committed
71
```
chenych's avatar
chenych committed
72

chenzk's avatar
v1.0  
chenzk committed
73
## 推理
chenych's avatar
chenych committed
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
推理框架有`transformers``vllm``SGLang`或者其他推理框架中任意一个即可,`单机单卡``单机多卡`章节根据模型大小自行选择即可。

### transformers
#### 单机推理
```
sh xxx.sh # 或python xxx.py
```

#### 多机推理
```
sh xxx.sh 或python xxx.py
```

### vllm
#### 单机推理
```
sh xxx.sh # 或python xxx.py
```

#### 多机推理
```
sh xxx.sh 或python xxx.py
```

### SGLang
#### 单机推理
```
sh xxx.sh # 或python xxx.py
```

#### 多机推理
chenzk's avatar
v1.0  
chenzk committed
105
```
chenzk's avatar
v1.0.1  
chenzk committed
106
sh xxx.sh 或python xxx.py
chenzk's avatar
v1.0  
chenzk committed
107
```
chenych's avatar
chenych committed
108
109
...

chenzk's avatar
v1.0  
chenzk committed
110
## result
chenzk's avatar
v1.0.6  
chenzk committed
111
此处填算法效果测试图(包括输入、输出)
liuhy's avatar
liuhy committed
112

chenzk's avatar
v1.0.6  
chenzk committed
113
114
115
<div align=center>
    <img src="./doc/xxx.png"/>
</div>
liuhy's avatar
liuhy committed
116

chenzk's avatar
v1.0.1  
chenzk committed
117
### 精度
liuhy's avatar
liuhy committed
118
119
测试数据:[test data](链接),使用的加速卡:xxx。

chenzk's avatar
v1.0  
chenzk committed
120
根据测试结果情况填写表格:
liuhy's avatar
liuhy committed
121
122
123
124
| xxx | xxx | xxx | xxx | xxx |
| :------: | :------: | :------: | :------: |:------: |
| xxx | xxx | xxx | xxx | xxx  |
| xxx | xx | xxx | xxx | xxx |
chenzk's avatar
v1.0.5  
chenzk committed
125

chenych's avatar
chenych committed
126
## 算法类别
chenzk's avatar
v1.0  
chenzk committed
127
`此处填算法类别`
chenych's avatar
chenych committed
128
填写此算法最主要的算法类别,数量为1,与icon图标类别一致,请勿随意命名。
chenzk's avatar
v1.0.7  
chenzk committed
129

chenzk's avatar
v1.0  
chenzk committed
130
## 预训练权重
chenych's avatar
chenych committed
131
132
133
134
|模型名称  | 权重大小  | DCU型号  | 最低卡数需求 |下载地址,填写公开预训练权重官网下载地址(必须),使用`[下载地址](链接)`格式,样例如下|
|:-----:|:----------:|:----------:|:----------:|:----------:|
|Qwen3 | 4B | K100AI,BW1000 | 1 | [下载地址](https://hf-mirror.com/Qwen/Qwen3-4B-Instruct-2507) |

135
## 源码仓库及问题反馈
chenzk's avatar
v1.0  
chenzk committed
136
- 此处填本项目gitlab地址
chenych's avatar
chenych committed
137

chenzk's avatar
v1.0  
chenzk committed
138
## 参考资料
chenzk's avatar
v1.0.1  
chenzk committed
139
- 此处填源github地址(方便使用者查看原github issue)
chenzk's avatar
v1.0  
chenzk committed
140
141
- 此处填参考项目或教程网址
- ......
chenzk's avatar
chenzk committed
142

chenzk's avatar
chenzk committed
143
`关于model.properties(必要)、LICENSE(必要)、CONTRIBUTORS、模型图标(必要)等其它信息提供参照: `[`ModelZooStd.md`](./ModelZooStd.md)
chenych's avatar
chenych committed
144
`各个模型需要保留原项目README.md,改名为README_origin.md即可。`