README.md 3.8 KB
Newer Older
wanglch's avatar
wanglch committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
# Kimi-VL-A3B-Instruct
## 论文

[Kimi-VL Technical Report](https://arxiv.org/abs/2504.07491)


## 模型结构

它的语言模型部分,采用了我们前一段时间发布的轻量级 MoE 模型 Moonlight,16B总参数,2.8B激活参数。视觉信息处理部分,采用了支持原生分辨率的 MoonViT 编码器——400M 参数 Vison Transformer 架构。该模型采用MoE语言模型、原生分辨率视觉编码器(MoonViT)和MLP投影仪。

<div align=center>
    <img src="./Pic/arch.png"/>
</div>

## 算法原理

ViT是Google团队提出的将Transformer应用在图像分类的模型。ViT将输入图片分为多个patch(16x16),再将每个patch投影为固定长度的向量送入Transformer,后续encoder的操作和原始Transformer中完全相同。但是因为对图片分类,因此在输入序列中加入一个特殊的token,该token对应的输出即为最后的类别预测

<div align=center>
    <img src="./Pic/theory.png"/>
</div>

## 环境配置
### Docker(方法一)
推荐使用docker方式运行, 此处提供[光源](https://www.sourcefind.cn/#/service-details)拉取docker镜像的地址与使用步骤
```
docker pull image.sourcefind.cn:5000/dcu/admin/base/pytorch:2.3.0-py3.10-dtk24.04.3-ubuntu20.04

docker run -it --shm-size=1024G -v /path/your_code_data/:/path/your_code_data/ -v /opt/hyhal:/opt/hyhal --network=host --privileged=true --device=/dev/kfd --device=/dev/dri/ --group-add video --name kimi-vl  <your IMAGE ID> bash # <your IMAGE ID>为以上拉取的docker的镜像ID替换

git clone http://developer.sourcefind.cn/codes/modelzoo/kimi-vl-a3b-instruct_pytorch.git

cd /path/your_code_data/

pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple

```
Tips:以上dtk驱动、python、torch等DCU相关工具版本需要严格一一对应。

### Dockerfile(方法二)
此处提供dockerfile的使用方法
```
wanglch's avatar
wanglch committed
43
44
git clone http://developer.sourcefind.cn/codes/modelzoo/kimi-vl-a3b-instruct_pytorch.git

wanglch's avatar
wanglch committed
45
46
47
48
49
50
51
52
53
54
55
56
docker build -t internvl:latest .
docker run --shm-size 500g --network=host --name=kimi-vl --privileged --device=/dev/kfd --device=/dev/dri --group-add video --cap-add=SYS_PTRACE --security-opt seccomp=unconfined -v 项目地址(绝对路径):/home/ -v /opt/hyhal:/opt/hyhal:ro -it <your IMAGE ID> bash


cd /path/your_code_data/

pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple

```
### Anaconda(方法三)
此处提供本地配置、编译的详细步骤,例如:

chenzk's avatar
chenzk committed
57
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。
wanglch's avatar
wanglch committed
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
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
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
```
DTK驱动:dtk24.04.3
python:3.10
torch:2.3.0
transformers>=4.48.0
```
`Tips:以上dtk驱动、python、torch等DCU相关工具版本需要严格一一对应`

其它非深度学习库参照requirement.txt安装:
```

git clone http://developer.sourcefind.cn/codes/modelzoo/kimi-vl-a3b-instruct_pytorch.git

cd /path/your_code_data/

pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple
```
## 数据集



## 训练



## 推理

### 单机单卡

```
python Kimi_VL_inference.py
```

### 单机多卡

```
CUDA_VISIBLE_DEVICES=0,1,2,3 python Kimi_VL_inference.py
```

## result


<div align=left>
    <img src="./Pic/result.png"/>
</div>

### 精度





## 应用场景

### 算法类别

`对话问答`

### 热点应用行业

`科研,教育,政府,金融`

## 预训练权重


HF/github下载地址为:[moonshotai/Kimi-VL-A3B-Instruct](https://huggingface.co/moonshotai/Kimi-VL-A3B-Instruct)

wanglch's avatar
wanglch committed
125
魔搭下载路径: [moonshotai/Kimi-VL-A3B-Instruct](https://www.modelscope.cn/models/moonshotai/Kimi-VL-A3B-Instruct/files)
wanglch's avatar
wanglch committed
126
127
128
129
130
131
132
133

## 源码仓库及问题反馈
-  http://developer.sourcefind.cn/codes/modelzoo/kimi-vl-a3b-instruct_pytorch.git

## 参考资料

- https://github.com/MoonshotAI/Kimi-VL