MinkowskiEngine简介

Minkowski引擎是一个用于稀疏张量的自动微分库。它支持所有标准神经网络层，例如对稀疏张量的卷积、池化和广播操作。

MinkowskiEngine安装

官方源码链接：GitHub - NVIDIA/MinkowskiEngine: Minkowski Engine is an auto-diff neural network library for high-dimensional sparse tensors

官方提供的安装教程只适合cuda-10.2和cuda-11.x，如果cuda版本≥12，可参考：MinkowskiEngine/installation_note.md at cuda12-installation · CiSong10/MinkowskiEngine · GitHub

我的cuda版本是12.4，参考“MinkowskiEngine/installation_note.md at cuda12-installation · CiSong10/MinkowskiEngine · GitHub”编译MinkowskiEngine时，仍然会报错。因此多安装了一个版本为12.1的cuda toolkit（CUDA Toolkit Archive | NVIDIA Developer），下载好cuda_12.1.0_530.30.02_linux.run后，执行以下命令安装cuda-12.1，

sudo sh cuda_12.1.0_530.30.02_linux.run

在安装界面中，使用空格键取消选择 "Driver"（驱动程序），只保留 "CUDA Toolkit 12.1" 被选中，因为NVIDIA驱动通常是向后兼容的，cuda 12.1工具包可以与较新的cuda-12.4驱动程序一起工作。安装程序会询问是否创建/usr/local/cuda符号链接，建议选择 "No"，这样不会更改当前默认的cuda版本。

修改~/.bashrc，在虚拟环境中指定使用cuda-12.1，

export CUDA_HOME=/usr/local/cuda-12.1
export LD_LIBRARY_PATH=$CUDA_HOME/lib64:$LD_LIBRARY_PATH

随后，验证cuda版本，

nvcc --version

"""
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2023 NVIDIA Corporation
Built on Tue_Feb__7_19:32:13_PST_2023
Cuda compilation tools, release 12.1, V12.1.66
Build cuda_12.1.r12.1/compiler.32415258_0
"""

确认cuda-12.1安装好后，再去安装其他相关库，

conda create -n <env_name> python=3.9
conda activate <env_name>
conda install -y pytorch==2.4.0 torchvision==0.19.0 torchaudio==2.4.0 pytorch-cuda=12.1 -c pytorch -c nvidia
conda install -y openblas-devel -c anaconda
conda install -y nvidia/label/cuda-12.1.0::cuda-toolkit

git clone https://github.com/CiSong10/MinkowskiEngine.git
cd MinkowskiEngine/
git checkout cuda12-installation

python setup.py install --blas=openblas 

安装好MinkowskiEngine后，在命令行会有如下提示，

执行以下命令可获取系统配置，

cd MinkowskiEngine/MinkowskiEngine
python diagnostics.py

我的配置如下，

==========System==========
Linux-5.15.0-133-generic-x86_64-with-glibc2.35
DISTRIB_ID=Ubuntu
DISTRIB_RELEASE=22.04
DISTRIB_CODENAME=jammy
DISTRIB_DESCRIPTION="Ubuntu 22.04.4 LTS"
3.9.21 (main, Dec 11 2024, 16:24:11) 
[GCC 11.2.0]
==========Pytorch==========
2.4.0
torch.cuda.is_available(): True
==========NVIDIA-SMI==========
/usr/bin/nvidia-smi
Driver Version 550.54.14
CUDA Version 12.4
VBIOS Version 95.02.3C.00.02
Image Version G002.0000.00.03
GSP Firmware Version 550.54.14
==========NVCC==========
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2023 NVIDIA Corporation
Built on Tue_Feb__7_19:32:13_PST_2023
Cuda compilation tools, release 12.1, V12.1.66
Build cuda_12.1.r12.1/compiler.32415258_0
==========CC==========
/usr/bin/c++
c++ (Ubuntu 9.5.0-1ubuntu1~22.04) 9.5.0
Copyright (C) 2019 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

==========MinkowskiEngine==========
0.5.4
MinkowskiEngine compiled with CUDA Support: True
NVCC version MinkowskiEngine is compiled: 12010
CUDART version MinkowskiEngine is compiled: 12010