Closed justanhduc closed 2 years ago
You can try to use SparseTensor
instead:
from torch_sparse import SparseTensor
sparse_mat = SparseTensor(row=..., col=..., value=..., size=(N, N))
out = sparse_mat @ dense_mat
Great it works! Thanks!
Hi @rusty1s. Thanks for making this repo. I tried to replace Pytorch
mm
byspmm
in my project but I got OOM. Is it already well-known or should the memory usage be similar somewhat?