torch.hsplit#

torch.hsplit(input,indices_or_sections)→ListofTensors#

Splitsinput, a tensor with one or more dimensions, into multiple tensorshorizontally according toindices_or_sections. Each split is a view ofinput.

Ifinput is one dimensional this is equivalent to callingtorch.tensor_split(input, indices_or_sections, dim=0) (the split dimension iszero), and ifinput has two or more dimensions it’s equivalent to callingtorch.tensor_split(input, indices_or_sections, dim=1) (the split dimension is 1),except that ifindices_or_sections is an integer it must evenly dividethe split dimension or a runtime error will be thrown.

This function is based on NumPy’snumpy.hsplit().

Parameters

• input (Tensor) – tensor to split.

• indices_or_sections (int orlist ortuple ofints) – See argument intorch.tensor_split().

Example:

>>>t=torch.arange(16.0).reshape(4,4)>>>ttensor([[ 0.,  1.,  2.,  3.],        [ 4.,  5.,  6.,  7.],        [ 8.,  9., 10., 11.],        [12., 13., 14., 15.]])>>>torch.hsplit(t,2)(tensor([[ 0.,  1.],         [ 4.,  5.],         [ 8.,  9.],         [12., 13.]]), tensor([[ 2.,  3.],         [ 6.,  7.],         [10., 11.],         [14., 15.]]))>>>torch.hsplit(t,[3,6])(tensor([[ 0.,  1.,  2.],         [ 4.,  5.,  6.],         [ 8.,  9., 10.],         [12., 13., 14.]]), tensor([[ 3.],         [ 7.],         [11.],         [15.]]), tensor([], size=(4, 0)))