Fast and generic implementation using OpenMP and CUDA

d-li14 / involution

[CVPR 2021] Involution: Inverting the Inherence of Convolution for Visual Recognition, a brand new neural operator

https://arxiv.org/abs/2103.06255

MIT License

1.31k stars 177 forks source link

Fast and generic implementation using OpenMP and CUDA #44

Open shikishima-TasakiLab opened 3 years ago

shikishima-TasakiLab commented 3 years ago

I have implemented a module using OpenMP and CUDA that runs faster while maintaining the memory efficiency of your CuPy implementation.

shikishima-TasakiLab/Involution-PyTorch

It also supports TorchScript and 16-bit float.

shikishima-TasakiLab/Involution-PyTorch#1

d-li14 commented 3 years ago

Great work! It will help a lot in practice! As I have mentioned in the README, would you please make a PR to contribute to this repo? Just to be on the safe side, I will run some experiments to double-check the reimplementation's correctness before merging it into the main branch. Thanks.

shikishima-TasakiLab commented 3 years ago

I made a PR. I did not merge the conflicting parts of the README, so please add module descriptions accordingly.

d-li14 commented 3 years ago

OK, I will verify and merge it as soon as I could.