[深度学习][原创]yolact编译DCNv2错误解决方法

项目地址：https://github.com/dbolya/yolact

使用yolact编译DCNv2提示

/media/howso/f/lu/myproj/yolact-master/external/DCNv2/src/cuda/dcn_v2_cuda.cu(107): error: identifier "THCState_getCurrentStream" is undefined

/media/howso/f/lu/myproj/yolact-master/external/DCNv2/src/cuda/dcn_v2_cuda.cu(279): error: identifier "THCState_getCurrentStream" is undefined

/media/howso/f/lu/myproj/yolact-master/external/DCNv2/src/cuda/dcn_v2_cuda.cu(324): error: identifier "THCudaBlas_Sgemv" is undefined

3 errors detected in the compilation of "/media/howso/f/lu/myproj/yolact-master/external/DCNv2/src/cuda/dcn_v2_cuda.cu".
error: command '/usr/local/cuda-11.0/bin/nvcc' failed with exit status 1
我的环境是：

ubuntu18.04

RTX3090

torch==1.7.0

torchvision==0.8.1

cuda11.0+cudnn8.1.1

原因是DCNv2对于RTX30系列支持不好，需要重新去https://github.com/MatthewHowe/DCNv2

下载，而且这样还不够需要在bash make.sh之前

export TORCH_CUDA_ARCH_LIST=7.5

虽然RTX3090算力是8.6但是还是会提示错误，设置7.5就可以