fn
conv1d_op
→TensorImplPtrint conv1d_op(const int & x, const int & W, const int & b, int stride_l, int pad_l, int dilation_l, int groups)Python wrappers
Public Python APIs implemented by this engine symbol.One-dimensional cross-correlation over a batch of signals.
Computes the autograd-aware op
where s is the stride and d is the dilation.
Parameters
xTensorImplPtrInput of shape
(B, C_in, L).WTensorImplPtrFilters of shape
(C_out, C_in / groups, KL).bTensorImplPtrBias of shape
(C_out,) or empty.stride_lint= NoneStride along the length axis. Default:
1.pad_lint= NoneZero-padding added to both ends of
L. Default: 0.dilation_lint= NoneKernel-element spacing along
L. Default: 1.groupsint= NoneChannel grouping. Default:
1.Returns
TensorImplPtrOutput of shape (B, C_out, L_out) where .