配置Stratified-Transformer、Point-Transformer系列

文章目录


conda install pytorch==1.7.1 torchvision==0.8.2 torchaudio==0.7.2 cudatoolkit=11.0 -c pytorch

Linux下切换cuda版本https://blog.csdn.net/Nirvana15/article/details/119062355

 
#写入切换函数
#*******   switch cuda
function _switch_cuda {
   v=$1
   export PATH=/usr/local/cuda-$v/bin:$PATH
   export CUDADIR=/usr/local/cuda-$v
   export CUDA_HOME=/usr/local/cuda-$v
   export LD_LIBRARY_PATH=/usr/local/cuda-$v/lib64:$LD_LIBRARY_PATH
   nvcc --version
}
_switch_cuda 11.0  #在此更改版本
#*******

ubuntu多版本gcc/g++切换
https://blog.csdn.net/weixin_43693967/article/details/123719406
手动配置update-alternatives选择gcc/g++版本

sudo update-alternatives --config gcc
sudo update-alternatives --config g++

ubuntu20.4安装gcc5.4
https://blog.csdn.net/xp_fangfei/article/details/123036267

import numpy as np

无法打开包括文件: “cuda_runtime_api.h”
配置正确的CUDA_HOME ,

$env:CUDA_HOME = "E:\USEAPP\CUDA111\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.1"
$env:CUDA_HOME = $env:CUDA_PATH_V11_1 # if CUDA_PATH_V11_1 is in envs:

或者

export CUDA_HOME = /usr/local/cuda-10.1

各版本CUDA 兼容的 GCC 版本总结
https://blog.csdn.net/u010087338/article/details/126649326

安装pytorch_gemometric.
https://rocstone.github.io/2021/04/23/%E5%AE%89%E8%A3%85pytorch-geometric%E6%8A%A5%E9%94%99AttributeError-NoneType-object-has-no-attribute-origin/

pip install torch-scatter --no-cache-dir
pip install torch-sparse --no-cache-dir
pip install torch-cluster --no-cache-dir
pip install torch-spline-conv --no-cache-dir
pip install torch-geometric 

你可能感兴趣的:(transformer,深度学习,人工智能)