I cann't compile compile the DCNv2

[nonlinear1]

I want to test the yolact++, but I cann't compile DCNv2. My enviroment is as follow:
OS: ubuntu18.04
cuda: 10.1
pytorch:1.5
When I compiled the DCNv2 I found the following errors:
/home/manifold/Dense_RGBD_VIO_Fusion/yolact/external/DCNv2/src/cuda/dcn_v2_cuda.cu(107): error: identifier "THCState_getCurrentStream" is undefined

/home/manifold/Dense_RGBD_VIO_Fusion/yolact/external/DCNv2/src/cuda/dcn_v2_cuda.cu(279): error: identifier "THCState_getCurrentStream" is undefined

2 errors detected in the compilation of "/tmp/tmpxft_00007b49_00000000-6_dcn_v2_cuda.cpp1.ii".
Could someone know how to solve this errors? Thank you very much

[starfish55555]

I also met, how do you solve it

[moganesyan]

I rolled back to torch 1.4.0 and torchvision 0.5.0 and solved the issue

[abhigoku10]

@moganesyan thanks for the solution will try it , same thing is happening in google colab

[moganesyan]

@abhigoku10 You're welcome. I was working on Colab as well, so should work for you too hopefully!

[khodabakhshih]

@moganesyan Thank you very much. It solved my issue in the google colab too.

[ahkarami]

is there any solution for use it on PyTorch 1.5?

[linkboyx]

This helps on Pytorch1.5. CharlesShang/DCNv2#58

[areebsyed]

I rolled back to torch 1.4.0 and torchvision 0.5.0 and solved the issue

Thankyou worked like a charm!

[blackCmd]

I rolled back to torch 1.4.0 and torchvision 0.5.0 and solved the issue

It works! Thank you

[lidar-nevermore]

this issuse can be solved by adding the following function:
cudaStream_t THCState_getCurrentStream(THCState *state) { return at::cuda::getCurrentCUDAStream().stream(); }
reference: https://github.com/pytorch/pytorch/pull/33376

[JiaWang0704]

this issuse can be solved by adding the following function:
cudaStream_t THCState_getCurrentStream(THCState *state) { return at::cuda::getCurrentCUDAStream().stream(); }
reference: https://github.com/pytorch/pytorch/pull/33376

hello,this resolve two errors but there still are three errors.as the following:(do you know the rearon?if you know, please tell me. thank u!)

[dungdo123]

I rolled back to torch 1.4.0 and torchvision 0.5.0 and solved the issue

Thanks, it solved my problem!

[UnderTheMangoTree]

I also met, how do you solve it

I raw back to pytorch 1.0.1, and the problem is resolved.

[DaBasile]

Hi guys I solved this problem with this issue https://github.com/CharlesShang/DCNv2/issues/57 replacing DCNv2 folder with the fork of the solution
my setup is
pytorch 1.10
cuda toolkit 11.4
cuda version 11.4

[Robotatron]

Hi guys I solved this problem with this issue https://github.com/CharlesShang/DCNv2/issues/57 replacing DCNv2 folder with the fork of the solution my setup is pytorch 1.10 cuda toolkit 11.4 cuda version 11.4

Your link does not work...