lucasjinreal / DCNv2_latest

DCNv2 supports decent pytorch such as torch 1.5+ (now 1.8+)
BSD 3-Clause "New" or "Revised" License
609 stars 125 forks source link

testcuda.py returns error but testcpu.py works #57

Open atabakd opened 2 years ago

atabakd commented 2 years ago

I am on the master branch and

# conda list |grep pytorch
ffmpeg                    4.3                  hf484d3e_0    pytorch
pytorch                   1.11.0          py3.7_cuda11.3_cudnn8.2.0_0    pytorch
pytorch-mutex             1.0                        cuda    pytorch
torchaudio                0.11.0               py37_cu113    pytorch
torchvision               0.12.0               py37_cu113    pytorch
# conda list |grep cu
cudatoolkit               11.3.1               h2bc3f7f_2  
ncurses                   6.3                  h7f8727e_2  
pytorch                   1.11.0          py3.7_cuda11.3_cudnn8.2.0_0    pytorch
pytorch-mutex             1.0                        cuda    pytorch
torchaudio                0.11.0               py37_cu113    pytorch
torchvision               0.12.0               py37_cu113    pytorch

But when I run the test, I get

# python testcuda.py
torch.Size([2, 64, 128, 128])                                                                                                           
torch.Size([20, 32, 7, 7])                                                                                                              
torch.Size([20, 32, 7, 7])                                                                                                              
torch.Size([20, 32, 7, 7])                                                                                                              
0.971507, 1.943014                                                                                                                      
0.971507, 1.943014                                                                                                                      
Zero offset failed  

However, the issue mentioned here gets Zero offset passed. Any idea?