Can the shape of convolutional kernels be designed by users themselves ？

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

1.19k stars 139 forks source link

Closed huohuohuohuohuohuohuohuo closed 9 months ago

huohuohuohuohuohuohuohuo commented 11 months ago

Can the shape of convolutional kernels be designed by users themselves ？

ys-2020 commented 11 months ago

Yes. Just specify the kernel shape (e.g., kernel_size = (1,3,3)) when defining the conv layer.