Update README_HIP.md

6a7901aa · quyuanhao123 · 4d28cb08 · 6a7901aa
Commit 6a7901aa authored Dec 22, 2023 by quyuanhao123
Hide whitespace changes
Inline Side-by-side

Showing with 75 additions and 63 deletions

README_HIP.md README_HIP.md +75 -63

No files found.
--- a/README_HIP.md
+++ b/README_HIP.md
+# <div align="center"><strong>torch-scatter-2.0.9</strong></div>
-**编译安装步骤**
+## 简介
+torch-scatter是一个在PyTorch库中使用的Python库，它用于从张量中随机选择元素并返回一个新的张量。这个库提供了一种简单的方法来创建具有随机标签的数据集，这对于许多机器学习任务非常有用，例如数据增强或生成对抗网络(GANs)。
- 1. 安装与dtk版本对应的pytorch torchvision whl（可在光合社区下载https://developer.hpccube.com/tool/）以及其他依赖库
+## 依赖安装
+ pytorch1.10或者pytorch1.13 以及对应的torchvision（建议dtk-22.04.2、dtk-23.04与dtk-23.10）
+ python 3.7-3.10
-    例如：
+### 1、使用源码编译方式安装
-    pip install https://cancon.hpccube.com:65024/directlink/4/pytorch/dtk22.10/torch-1.10.0a0+git2040069.dtk2210-cp38-cp38-manylinux2014_x86_64.whl
+#### 编译环境准备
-    pip install https://cancon.hpccube.com:65024/directlink/4/vision/dtk22.10/torchvision-0.10.0a0+e04d001.dtk2210-cp38-cp38-manylinux2014_x86_64.whl
+提供2种环境准备方式：
-    pip install -r requirements.txt
+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. 添加编译时的conda环境及部分库的环境变量
+2. 基于现有python环境：安装pytorch和torchvision，whl包下载目录：[https://cancon.hpccube.com:65024/4/main/pytorch](https://cancon.hpccube.com:65024/4/main/pytorch)，[https://cancon.hpccube.com:65024/4/main/vision](https://cancon.hpccube.com:65024/4/main/vision)，根据python、dtk版本,下载对应pytorch和torchvision的whl包。安装命令如下：
+```shell
-    - 2.1 激活对应的conda环境：
+pip install torch* (下载的torch的whl包)
+pip install torchvision* (下载的torchvision的whl包)
+pip install setuptools==59.5.0 wheel
-      source ~/miniconda3/etc/profile.d/conda.sh
+```
-      conda activate torch1.10_py39_dtk22.10
+#### 源码编译安装
+- 代码下载
-    - 2.2 加载对应的module，包括dtk：
+```shell
+git clone http://developer.hpccube.com/codes/aicomponent/torch-scatter # 根据编译需要切换分支
+```
-      module purge
+- 源码编译（进入torch-scatter目录）：
+```
-      module load compiler/devtoolset/7.3.1 mpi/hpcx/gcc-7.3.1 compiler/dtk/22.10.1
+export C_INCLUDE_PATH=/public/software/apps/DeepLearning/PyTorch_Lib/gflags-2.1.2-build/include:$C_INCLUDE_PATH
+export CPLUS_INCLUDE_PATH=/public/software/apps/DeepLearning/PyTorch_Lib/gflags-2.1.2-build/include:$CPLUS_INCLUDE_PATH
-      module list
+export C_INCLUDE_PATH=/public/software/apps/DeepLearning/PyTorch_Lib/glog-build/include:$C_INCLUDE_PATH
+export CPLUS_INCLUDE_PATH=/public/software/apps/DeepLearning/PyTorch_Lib/glog-build/include:$CPLUS_INCLUDE_PATH
-    - 2.3 加载所需的依赖库的环境变量（根据各集群实际路径调整）：
+export C_INCLUDE_PATH=$ROCM_PATH/rocrand/include:$C_INCLUDE_PATH
+export CPLUS_INCLUDE_PATH=$ROCM_PATH/rocrand/include:$CPLUS_INCLUDE_PATH
+export LD_LIBRARY_PATH=$ROCM_PATH/rocrand/lib:$LD_LIBRARY_PATH
-      export C_INCLUDE_PATH=/public/software/apps/DeepLearning/PyTorch_Lib/gflags-2.1.2-build/include:$C_INCLUDE_PATH
+export FORCE_ONLY_HIP=1
+export CC=hipcc
-      export CPLUS_INCLUDE_PATH=/public/software/apps/DeepLearning/PyTorch_Lib/gflags-2.1.2-build/include:$CPLUS_INCLUDE_PATH
+export CXX=hipcc
-      export C_INCLUDE_PATH=/public/software/apps/DeepLearning/PyTorch_Lib/glog-build/include:$C_INCLUDE_PATH
+python setup.py install
+```
-      export CPLUS_INCLUDE_PATH=/public/software/apps/DeepLearning/PyTorch_Lib/glog-build/include:$CPLUS_INCLUDE_PATH
+#### 注意事项
+ 若使用pip install下载安装过慢，可添加pypi清华源：-i https://pypi.tuna.tsinghua.edu.cn/simple/
-      export C_INCLUDE_PATH=$ROCM_PATH/rocrand/include:$C_INCLUDE_PATH
+ ROCM_PATH为dtk的路径，默认为/opt/dtk
-      export CPLUS_INCLUDE_PATH=$ROCM_PATH/rocrand/include:$CPLUS_INCLUDE_PATH
+## 验证
-      export LD_LIBRARY_PATH=$ROCM_PATH/rocrand/lib:$LD_LIBRARY_PATH
+```python
+import torch
-    - 2.4 修改编译器环境变量：
+from torch_scatter import scatter_max
+src = torch.tensor([[2, 0, 1, 4, 3], [0, 2, 1, 3, 4]])
-      export FORCE_ONLY_HIP=1
+index = torch.tensor([[4, 5, 4, 2, 3], [0, 0, 2, 2, 1]])
-      export CC=hipcc
+out, argmax = scatter_max(src, index, dim=-1)
+```
-      export CXX=hipcc
+```
- 3. 编译安装
+print(out)
+tensor([[0, 0, 4, 3, 2, 0],
+        [2, 4, 3, 0, 0, 0]])
-      python setup.py install
+print(argmax)
+tensor([[5, 5, 3, 4, 0, 1]
+        [1, 4, 3, 5, 5, 5]])
+```
+## Known Issue
+- 该库没有基于cpu环境修改，仅支持dcu，请在有dcu卡的环境运行。
+- 如需完整使用所有pyg功能，请pip install torch-geometric
+## 参考资料
+- [README_ORIGIN](README_ORIGIN.md)
+- [https://pypi.org/project/torch-scatter/2.0.9/](https://pypi.org/project/torch-scatter/2.0.9/)