README.md 2.33 KB
Newer Older
zhuwenwen's avatar
zhuwenwen committed
1
2
3
# <div align="center"><strong>FastFold</strong></div>
## 简介
FastFold是蛋白质结构预测模型推理的库,主要提供以下功能:蛋白质单体和多体的推理,提供Evoformer的高性能实现,支持多卡分布式推理。FastFold官方github地址:[https://github.com/hpcaitech/FastFold](https://github.com/hpcaitech/FastFold)
Shenggan's avatar
Shenggan committed
4

zhuwenwen's avatar
zhuwenwen committed
5
6
7
## 安装
安装FastFold,你需要
+ Python 3.8 或者 3.9.
Shenggan's avatar
Shenggan committed
8

zhuwenwen's avatar
zhuwenwen committed
9
### 使用pip方式安装
zhuwenwen's avatar
zhuwenwen committed
10
11
fastfold whl包下载目录:[https://cancon.hpccube.com:65024/4/main/fastfold/dtk23.04](https://cancon.hpccube.com:65024/4/main/fastfold/dtk23.04)
根据对应的pytorch版本和python版本,下载对应fastfold的whl包
Shenggan's avatar
Shenggan committed
12

Fazzie-Maqianli's avatar
Fazzie-Maqianli committed
13
```shell
zhuwenwen's avatar
zhuwenwen committed
14
pip install fastfold* (下载的fastfold的whl包)
Shenggan's avatar
Shenggan committed
15
16
```

zhuwenwen's avatar
zhuwenwen committed
17
### 使用源码编译方式安装
18

zhuwenwen's avatar
zhuwenwen committed
19
#### 编译环境准备
zhuwenwen's avatar
zhuwenwen committed
20
21
22
23
24
提供2种环境准备方式:

1. 基于光源pytorch基础镜像环境:镜像下载地址:[https://sourcefind.cn/#/image/dcu/pytorch](https://sourcefind.cn/#/image/dcu/pytorch),根据pytorch、python、dtk及系统下载对应的镜像版本。

2. 基于现有python环境:安装pytorch,pytorch whl包下载目录:[https://cancon.hpccube.com:65024/4/main/pytorch/dtk23.04](https://cancon.hpccube.com:65024/4/main/pytorch/dtk23.04),根据python、dtk版本,下载对应pytorch的whl包。安装命令如下:
Shenggan's avatar
Shenggan committed
25
```shell
zhuwenwen's avatar
zhuwenwen committed
26
pip install torch* (下载的torch的whl包)
zhuwenwen's avatar
zhuwenwen committed
27
pip install setuptools==59.5.0 wheel
28
29
```

zhuwenwen's avatar
zhuwenwen committed
30
31
#### 源码编译安装
- 代码下载
32
```shell
zhuwenwen's avatar
zhuwenwen committed
33
git clone https://developer.hpccube.com/codes/aicomponent/fastfold  # 根据需要的分支进行切换
34
35
```

zhuwenwen's avatar
zhuwenwen committed
36
37
38
- 提供2种源码编译方式(进入fastfold目录):
```
1. 编译whl包并安装
zhuwenwen's avatar
zhuwenwen committed
39
export CPLUS_INCLUDE_PATH=$ROCM_PATH/include:$ROCM_PATH/cuda/targets/x86_64-linux/include:$CPLUS_INCLUDE_PATH
zhuwenwen's avatar
zhuwenwen committed
40
python setup.py bdist_wheel 
zhuwenwen's avatar
zhuwenwen committed
41
pip install dist/fastfold*
zhuwenwen's avatar
zhuwenwen committed
42

zhuwenwen's avatar
zhuwenwen committed
43
44
45
2. 源码编译安装
export CPLUS_INCLUDE_PATH=$ROCM_PATH/include:$ROCM_PATH/cuda/targets/x86_64-linux/include:$CPLUS_INCLUDE_PATH
python3 setup.py install 
Shenggan's avatar
Shenggan committed
46
47
```

zhuwenwen's avatar
zhuwenwen committed
48
## 版本号查询
zhuwenwen's avatar
zhuwenwen committed
49
- python -c "import fastfold; print(fastfold.\_\_version__)",版本号与官方版本同步,查询该软件的版本号,例如0.2.0;
zhuwenwen's avatar
zhuwenwen committed
50

zhuwenwen's avatar
zhuwenwen committed
51
52
## Known Issue
-
zhuwenwen's avatar
zhuwenwen committed
53

zhuwenwen's avatar
zhuwenwen committed
54
55
56
## Note
+ 若使用 pip install 下载安装过慢,可添加源:-i https://pypi.tuna.tsinghua.edu.cn/simple/
+ ROCM_PATH为dtk的路径,默认为/opt/dtkxxx
Shenggan's avatar
Shenggan committed
57

zhuwenwen's avatar
zhuwenwen committed
58
## 参考
zhuwenwen's avatar
zhuwenwen committed
59
- [README_ORIGIN](README_ORIGIN.md)