search

tkipf / pygcn

Graph Convolutional Networks in PyTorch
MIT License
5.16k stars 1.23k forks source link

torch.spmm #19

Open zhuchangjiang opened 6 years ago

zhuchangjiang commented 6 years ago

what does this mean?

tkipf commented 6 years ago

Sparse matrix multiplication

On Mon 24. Sep 2018 at 11:12 changjiang.zhu notifications@github.com wrote:

what does this mean?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/tkipf/pygcn/issues/19#issuecomment-423929073, or mute the thread https://github.com/notifications/unsubscribe-auth/AHAcYOikqTu-yGmoQqFdfsm3xDQsJzhOks5ueLAMgaJpZM4W2W9g .

Tanyjiang commented 5 years ago

It does not work in Pytorch 0.4.1. This function only supports the torch.tensor but not Variable.

tkipf commented 5 years ago

Can you try PyTorch 0.5?

Tanyjiang commented 5 years ago

I tried a new defined layer for torch.spmm(). It works. Thanks

jgbradley1 commented 5 years ago

As a note, torch.spmm may move to torch.sparse.spmm. The API states it is experimental at the moment.

InstantWindy commented 5 years ago

I tried a new defined layer for torch.spmm(). It works. Thanks @Tanyjiang Hey! How do you define the 'spmm' layer? Thanks !

kprojiang commented 3 years ago

In PyTorch 1.8, torch.spmm is deprecated. Unify matrix multiplications operations Deprecate spmm and dsmm functions TORCH.SPARSE 1.8

As a note, torch.spmm may move to torch.sparse.spmm. The API states it is experimental at the moment.

Cogito2012 commented 3 years ago

Could anyone provide a substitution for the deprecated torch.spmm?

salcc commented 1 year ago

torch.mm should work, as torch.spmm was just an alias for it, as mentioned in https://github.com/pytorch/pytorch/issues/46164.