"...text-generation-inference.git" did not exist on "762dbf3f198a9fbe7edffd60edc909e60e66878a"
README.md 2.65 KB
Newer Older
zzg_666's avatar
zzg_666 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
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
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
# VibeThinker
## 论文
[VibeThinker](https://github.com/WeiboAI/VibeThinker/blob/main/VibeThinker-1.5B.pdf)

## 模型简介
VibeThinker-1.5B 是一个拥有15亿参数的密集型语言模型。其总训练成本仅为7800美元,却在推理性能上与更大的模型如GPT OSS-20B Medium相当。

<div align=center>
    <img src="./doc/bench0.png"/>
</div>

**关键性能数据:**

- **数学推理:** 在三大数学基准AIME24、AIME25和HMMT25上,它的得分分别为80.3、74.4和50.4,均超过了参数量超过其400倍的初始DeepSeek R1模型(得分分别为79.8、70.0和41.7)。
- **代码生成:** 它在LiveCodeBench v5上的得分为55.9,在v6上的得分为51.1。其v6得分略高于Magistral Medium(50.3),突显了其强大的推理性能。

<div align=center>
    <img src="./doc/bench1.png"/>
</div>



## 环境依赖

| 软件 | 版本 |
| :------: | :------: |
| DTK | 25.04.2 |
| python | 3.10.12 |
| transformers | 4.57.1 |
| flash-attn | 2.6.1+das.opt1.dtk2504 |
| torch | 2.7.1+das.opt1.dtk25042 |
| triton | 3.1+das.opt1.3c5d12d.dtk25041 |


当前仅支持镜像:
- 挂载地址`-v`根据实际模型情况修改

```bash
docker pull image.sourcefind.cn:5000/dcu/admin/base/pytorch:2.7.1-ubuntu22.04-dtk25.04.2-py3.10-alpha
docker run -it --shm-size 60g --network=host --name vibethinker --privileged --device=/dev/kfd --device=/dev/dri --device=/dev/mkfd --group-add video --cap-add=SYS_PTRACE --security-opt seccomp=unconfined -u root -v /opt/hyhal/:/opt/hyhal/:ro -v /path/your_code_path/:/path/your_code_path/ image.sourcefind.cn:5000/dcu/admin/base/pytorch:2.7.1-ubuntu22.04-dtk25.04.2-py3.10-alpha bash
pip install transformers>=4.54.0
pip install accelerate
cd whl
pip install flash_attn*.whl
```
更多镜像可前往[光源](https://sourcefind.cn/#/service-list)下载使用。

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。

## 数据集
暂无

## 训练
暂无

## 推理


### transformers
#### 单机推理
```bash
export HIP_VISIBLE_DEVICES=0
python VibeThinker.py
```

## 效果展示
<div align=center>
    <img src="./doc/result.png"/>
</div>

### 精度
DCU与GPU精度一致,推理框架:pytorch。

## 预训练权重
| 模型名称  | 权重大小  | DCU型号  | 最低卡数需求 |下载地址|
|:-----:|:----------:|:----------:|:---------------------:|:----------:|
| VibeThinker-1.5B | 1.5B | K100AI | 1 | [下载地址](https://huggingface.co/WeiboAI/VibeThinker-1.5B) |

## 源码仓库及问题反馈
- https://developer.sourcefind.cn/codes/modelzoo/vibethinker-pytorch

## 参考资料
- https://github.com/WeiboAI/VibeThinker