README-zh-Hans.md 8.23 KB
Newer Older
binmakeswell's avatar
binmakeswell committed
1
# Colossal-AI
2
<div id="top" align="center">
binmakeswell's avatar
binmakeswell committed
3

4
5
6
   [![logo](https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/Colossal-AI_logo.png)](https://www.colossalai.org/)

   一个整合高效并行技术的AI大模型训练系统。
binmakeswell's avatar
binmakeswell committed
7
8
9

   <h3> <a href="https://arxiv.org/abs/2110.14883"> 论文 </a> | 
   <a href="https://www.colossalai.org/"> 文档 </a> | 
10
   <a href="https://github.com/hpcaitech/ColossalAI-Examples"> 例程 </a> |   
binmakeswell's avatar
binmakeswell committed
11
   <a href="https://github.com/hpcaitech/ColossalAI/discussions"> 论坛 </a> | 
12
   <a href="https://medium.com/@hpcaitech"> 博客 </a></h3>
binmakeswell's avatar
binmakeswell committed
13

Frank Lee's avatar
Frank Lee committed
14
   [![Build](https://github.com/hpcaitech/ColossalAI/actions/workflows/build.yml/badge.svg)](https://github.com/hpcaitech/ColossalAI/actions/workflows/build.yml)
binmakeswell's avatar
binmakeswell committed
15
   [![Documentation](https://readthedocs.org/projects/colossalai/badge/?version=latest)](https://colossalai.readthedocs.io/en/latest/?badge=latest)
16
   [![CodeFactor](https://www.codefactor.io/repository/github/hpcaitech/colossalai/badge)](https://www.codefactor.io/repository/github/hpcaitech/colossalai)
Frank Lee's avatar
Frank Lee committed
17
   [![HuggingFace badge](https://img.shields.io/badge/%F0%9F%A4%97HuggingFace-Join-yellow)](https://huggingface.co/hpcai-tech)
binmakeswell's avatar
binmakeswell committed
18
   [![slack badge](https://img.shields.io/badge/Slack-join-blueviolet?logo=slack&amp)](https://join.slack.com/t/colossalaiworkspace/shared_invite/zt-z7b26eeb-CBp7jouvu~r0~lcFzX832w)
19
   [![WeChat badge](https://img.shields.io/badge/微信-加入-green?logo=wechat&amp)](https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/WeChat.png)
binmakeswell's avatar
binmakeswell committed
20
21

   | [English](README.md) | [中文](README-zh-Hans.md) |
22

binmakeswell's avatar
binmakeswell committed
23
</div>
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


## 目录
<ul>
 <li><a href="#特点">特点</a> </li>
 <li>
   <a href="#展示样例">展示样例</a> 
   <ul>
     <li><a href="#ViT">ViT</a></li>
     <li><a href="#GPT-3">GPT-3</a></li>
     <li><a href="#GPT-2">GPT-2</a></li>
     <li><a href="#BERT">BERT</a></li>
   </ul>
 </li>

 <li>
   <a href="#安装">安装</a>
   <ul>
     <li><a href="#PyPI">PyPI</a></li>
     <li><a href="#从源代码安装">从源代码安装</a></li>
   </ul>
 </li>
 <li><a href="#使用-Docker">使用 Docker</a></li>
 <li><a href="#社区">社区</a></li>
 <li><a href="#做出贡献">做出贡献</a></li>
 <li><a href="#快速预览">快速预览</a></li>
   <ul>
     <li><a href="#几行代码开启分布式训练">几行代码开启分布式训练</a></li>
     <li><a href="#构建一个简单的2维并行模型">构建一个简单的2维并行模型</a></li>
   </ul>
 <li><a href="#引用我们">引用我们</a></li>
</ul>
binmakeswell's avatar
binmakeswell committed
56
57
58
59
60
61
62
63
64
65
66
67
68
69

## 特点

Colossal-AI为您提供了一系列并行训练组件。我们的目标是让您的分布式AI模型训练像普通的单GPU模型一样简单。我们提供的友好工具可以让您在几行代码内快速开始分布式训练。

- 数据并行
- 流水线并行
- 1维, 2维, 2.5维, 3维张量并行 
- 序列并行
- 友好的trainer和engine
- 可扩展新的并行方式
- 混合精度
- 零冗余优化器 (ZeRO)

70
71
72
<p align="right">(<a href="#top">返回顶端</a>)</p>

## 展示样例
binmakeswell's avatar
binmakeswell committed
73
74
### ViT

75
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/ViT.png" width="450" />
76

77
- 14倍批大小和5倍训练速度(张量并行=64)
binmakeswell's avatar
binmakeswell committed
78

79
### GPT-3
80
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/GPT3.png" width=700/>
binmakeswell's avatar
binmakeswell committed
81

82
- 释放 50% GPU 资源占用, 或 10.7% 加速
83
84

### GPT-2
85
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/GPT2.png" width=800/>
86

87
88
89
90
91
- 降低11倍GPU显存占用,或超线性扩展(张量并行)

<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/Colossal-AI%20with%20ZeRO.jpg" width=393>

- 能训练接近11倍大小的模型(ZeRO) 
binmakeswell's avatar
binmakeswell committed
92
93

### BERT
94
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/BERT.png" width=800/>
binmakeswell's avatar
binmakeswell committed
95

96
- 2倍训练速度,或1.5倍序列长度
binmakeswell's avatar
binmakeswell committed
97
98
99

请访问我们的[文档和教程](https://www.colossalai.org/)以了解详情。

100
<p align="right">(<a href="#top">返回顶端</a>)</p>
binmakeswell's avatar
binmakeswell committed
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140

## 安装

### PyPI

```bash
pip install colossalai
```
该命令将会安装CUDA extension,如果你已安装CUDA, NVCC和torch。 

如果你不想安装CUDA extension, 可在命令中添加`--global-option="--no_cuda_ext"`, 例如:
```bash
pip install colossalai --global-option="--no_cuda_ext"
```

如果你想使用`ZeRO`, 你可以使用:
```bash
pip install colossalai[zero]
```

### 从源代码安装

> Colossal-AI的版本将与该项目的主分支保持一致。欢迎通过issue反馈你遇到的任何问题 :)

```shell
git clone https://github.com/hpcaitech/ColossalAI.git
cd ColossalAI
# 安装依赖
pip install -r requirements/requirements.txt

# 安装 colossalai
pip install .
```

如果你不想安装和使用CUDA kernel fusion (使用fused优化器需安装):

```shell
pip install --global-option="--no_cuda_ext" .
```

141
142
<p align="right">(<a href="#top">返回顶端</a>)</p>

binmakeswell's avatar
binmakeswell committed
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
## 使用 Docker

运行以下命令从我们提供的docker文件中建立docker镜像。

```bash
cd ColossalAI
docker build -t colossalai ./docker
```

运行以下命令从以交互式启动docker镜像.

```bash
docker run -ti --gpus all --rm --ipc=host colossalai bash
```

158
<p align="right">(<a href="#top">返回顶端</a>)</p>
binmakeswell's avatar
binmakeswell committed
159
160
161
162

## 社区
欢迎通过[论坛](https://github.com/hpcaitech/ColossalAI/discussions),
[Slack](https://join.slack.com/t/colossalaiworkspace/shared_invite/zt-z7b26eeb-CBp7jouvu~r0~lcFzX832w),
163
[微信](https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/WeChat.png "qrcode")加入Colossal-AI社区,与我们分享你的建议和问题。
binmakeswell's avatar
binmakeswell committed
164
165


binmakeswell's avatar
binmakeswell committed
166
167
168
169
## 做出贡献

欢迎为该项目做出贡献,请参阅[贡献指南](./CONTRIBUTING.md)

binmakeswell's avatar
binmakeswell committed
170
171
172
173
174
175
真诚感谢所有贡献者!

<a href="https://github.com/hpcaitech/ColossalAI/graphs/contributors"><img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/contributor_avatar.png" width="800px"></a>

*贡献者头像的展示顺序是随机的。*

176
<p align="right">(<a href="#top">返回顶端</a>)</p>
binmakeswell's avatar
binmakeswell committed
177
178
179

## 快速预览

180
### 几行代码开启分布式训练
binmakeswell's avatar
binmakeswell committed
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256

```python
import colossalai
from colossalai.utils import get_dataloader


# my_config可以是config文件的路径或字典对象
# 'localhost' 仅适用于单节点,在多节点时需指明节点名
colossalai.launch(
    config=my_config,
    rank=rank,
    world_size=world_size,
    backend='nccl',
    port=29500,
    host='localhost'
)

# 构建模型
model = ...

# 构建数据集, dataloader会默认处理分布式数据sampler
train_dataset = ...
train_dataloader = get_dataloader(dataset=dataset,
                                shuffle=True
                                )


# 构建优化器
optimizer = ...

# 构建损失函数
criterion = ...

# 初始化colossalai
engine, train_dataloader, _, _ = colossalai.initialize(
    model=model,
    optimizer=optimizer,
    criterion=criterion,
    train_dataloader=train_dataloader
)

# 开始训练
engine.train()
for epoch in range(NUM_EPOCHS):
    for data, label in train_dataloader:
        engine.zero_grad()
        output = engine(data)
        loss = engine.criterion(output, label)
        engine.backward(loss)
        engine.step()

```

### 构建一个简单的2维并行模型

假设我们有一个非常巨大的MLP模型,它巨大的hidden size使得它难以被单个GPU容纳。我们可以将该模型的权重以二维网格的形式分配到多个GPU上,且保持你熟悉的模型构建方式。

```python
from colossalai.nn import Linear2D
import torch.nn as nn


class MLP_2D(nn.Module):

    def __init__(self):
        super().__init__()
        self.linear_1 = Linear2D(in_features=1024, out_features=16384)
        self.linear_2 = Linear2D(in_features=16384, out_features=1024)

    def forward(self, x):
        x = self.linear_1(x)
        x = self.linear_2(x)
        return x

```

257
<p align="right">(<a href="#top">返回顶端</a>)</p>
binmakeswell's avatar
binmakeswell committed
258

259
## 引用我们
binmakeswell's avatar
binmakeswell committed
260
261
262
263
264
265
266
267
268

```
@article{bian2021colossal,
  title={Colossal-AI: A Unified Deep Learning System For Large-Scale Parallel Training},
  author={Bian, Zhengda and Liu, Hongxin and Wang, Boxiang and Huang, Haichen and Li, Yongbin and Wang, Chuanrui and Cui, Fan and You, Yang},
  journal={arXiv preprint arXiv:2110.14883},
  year={2021}
}
```
269
270

<p align="right">(<a href="#top">返回顶端</a>)</p>