You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hey Yan yan,
Thank you for the amazing work.
Considering that SparseConvNet has more interface functions, such as average pooling and concatTable ect., and that only differences between spconv and SparseConvNet are gpu indice generation and gather-gemm-scatter algorithm, are you planning to add gpu indice generation and gather-gemm-scatter into SparseConvNet released by Facebook or just to develop more functions based on spconv?
The text was updated successfully, but these errors were encountered:
I have no plan to merge to SparseConvNet. My python and c++ code both are completely different with SparseConvNet.
The avg pool seems not used in modern convolution models. The ConcatTable is just a python module. Many layers in SparseConvNet can be implemented in pure-pytorch.
Hey Yan yan,
Thank you for the amazing work.
Considering that SparseConvNet has more interface functions, such as average pooling and concatTable ect., and that only differences between spconv and SparseConvNet are gpu indice generation and gather-gemm-scatter algorithm, are you planning to add gpu indice generation and gather-gemm-scatter into SparseConvNet released by Facebook or just to develop more functions based on spconv?
The text was updated successfully, but these errors were encountered: