mit-han-lab / torchsparse

[MICRO'23, MLSys'22] TorchSparse: Efficient Training and Inference Framework for Sparse Convolution on GPUs.
https://torchsparse.mit.edu
MIT License
1.22k stars 143 forks source link

How can I imitate spconv.Conv3d with stride = 1? #333

Open chriskeraly opened 1 month ago

chriskeraly commented 1 month ago

As far as I understand, this seems impossible since if stride =1 I directly get sparse submanifold behavior? Is there a work around?