Not Compatible with PyTorch 1.7 + CUDA 11

baowenbo / DAIN

Depth-Aware Video Frame Interpolation (CVPR 2019)

https://sites.google.com/view/wenbobao/dain

MIT License

8.19k stars 840 forks source link

Not Compatible with PyTorch 1.7 + CUDA 11 #118

Open markyang59 opened 3 years ago

markyang59 commented 3 years ago

Dear It seems this code is not compatible with PyTorch 1.7 (lateset now) Don't have plan for 1.7

iBobbyTS commented 3 years ago

The latest version or PyTorch you can use for DAIN is 1.4.0, after that, you need a static method for the forward functions, of all the networks. So it's a huge amount of work. For CUDA 11, did you try if pip install torch==1.4.0 works? If not, you can refer to PyTorch/PyTorch and compile torch 1.4.0 with CUDA 11 by your self.

flobauer commented 3 years ago

I am currently struggling getting DAIN to work on ubuntu 20.04, gtx3080 and CUDA 11. The mix of getting pytorch 1.4 work with CUDA 11 probably takes me longer than it should.

laomao0 commented 2 years ago

If you do not want to build CUDA programs. We provide the CUPY version of those packages. The cupy files do not need to be built. please refer to: https://github.com/laomao0/cupy_packages