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

Fixed GroupNorm #245

Closed matvogel closed 9 months ago

matvogel commented 1 year ago

Fixed the GroupNorm module detection of batch size because the bach index is not at index 0 of the coordinate vector.

ys-2020 commented 9 months ago

Hi, we have fixed this problem in previous commits. Thank you very much for your efforts and interest in TorchSparse.