ubuntu20.04 cuda11.4 cudnn8.2 编译、使用libtorch1.8 1.9

最新推荐文章于 2024-06-29 11:44:46 发布

细雨火锅

最新推荐文章于 2024-06-29 11:44:46 发布

阅读量1.3k

点赞数

本文链接：https://blog.csdn.net/kang0529wei/article/details/119237942

版权

本文讲述了在新环境中编译libtorch遇到的问题，包括去掉CAFFE2选项、处理GPU不兼容错误、链接库问题，以及解决dlopen undefined reference。通过详细步骤和代码实例，帮助读者顺利部署GPU版本的libtorch。

摘要由CSDN通过智能技术生成

一不小心环境弄得比较新，坑比较多。

主要编译libtorch。

gcc 9.3.0

g++ 9.3.0

主要使用4个库

libc10.so
lib/libc10_cuda.so
libtorch_cpu.so
libtorch_cuda.so

主要问题:

一、需要注意的一个地方在CMakeList.txt 去掉BUILD_CAFFE2的选项，使用官方的命令，可以编译通过。

export CMAKE_PREFIX_PATH=${CONDA_PREFIX:-"$(dirname $(which conda))/../"}
python setup.py build --cmake-only
ccmake build  # or cmake-gui build

二、使用的时候CPU没问题，GPU出现：

判断可用设备为0

terminate called after throwing an instance of 'c10::NotImplementedError'
what(): Could not run 'aten::empty_strided' with arguments from the 'CUDA' backend. This could be because the operator doesn't exist for this backend, or was omitted during the selective/custom build process (if using custom build). If you are a Facebook employee using PyTorch on mobile, please visit https://fburl.com/ptmfixes for possible resolutions. 'aten::empty_strided' is only available for these backends: [CPU, Meta, BackendSelect, Named, ADInplaceOrView, AutogradOther, AutogradCPU, AutogradCUDA, AutogradXLA, UNKNOWN_TENSOR_TYPE_ID, AutogradMLC, AutogradHPU, AutogradNestedTensor, AutogradPrivateUse1, AutogradPrivateUse2, AutogradPrivateUse3, Tracer, Autocast, Batched, VmapMode].

解决方法

在使用GPU之前加入这一行代码，可以正常使用

try {
void *handle = dlopen("/usr/local/libtorch1.9/lib/libtorch_cuda.so",RTLD_LAZY);
std::cout << torch::cuda::is_available() << std::endl;
torch::Tensor tensor = at::tensor({ -1, 1 }, at::kCUDA);
}catch (exception& ex) {
std::cout << ex.what() << std::endl;
}

三、/usr/bin/ld: CMakeFiles/example-app.dir/example-app.o: undefined reference to symbol 'dlopen@@GLIBC_2.2.5'
/usr/bin/ld: /lib/x86_64-linux-gnu/libdl.so.2: error adding symbols: DSO missing from command line
collect2: error: ld returned 1 exit status
make[2]: *** [CMakeFiles/example-app.dir/build.make:102：example-app] 错误 1
make[1]: *** [CMakeFiles/Makefile2:76：CMakeFiles/example-app.dir/all] 错误 2
make: *** [Makefile:84：all] 错误 2

在CMakeList.txt 中加入

target_link_libraries(example-app "/usr/local/libtorch1.9/lib/libc10.so" ${CMAKE_DL_LIBS})
target_link_libraries(example-app "/usr/local/libtorch1.9/lib/libc10_cuda.so" ${CMAKE_DL_LIBS})
target_link_libraries(example-app "/usr/local/libtorch1.9/lib/libtorch_cpu.so" ${CMAKE_DL_LIBS})
target_link_libraries(example-app "/usr/local/libtorch1.9/lib/libtorch_cuda.so" ${CMAKE_DL_LIBS})