Appying Conv1d to sparce tensor

Hello, I am trying to implement CDSSM in PyTorch. I want to use sparse tensor to represent input data. In my model sparse tensor should be passed to Conv1d layer, but when I am trying to do it, I receives exception:

RuntimeError: THCSTensor::rawStrides() not supported

What should I do to use sparse tensors with Conv1d layer and is it even possible?

It’s not supported right now. You could think of a convolution as a matrix multiply and try to do a sparse x dense matrix multiply.