oneflow.hsplit

oneflow.hsplit(input, indices_or_sections)List of Tensors

The documentation is referenced from: https://pytorch.org/docs/1.10/generated/torch.hsplit.html.

Splits input, a tensor with one or more dimensions, into multiple tensors horizontally according to indices_or_sections. Each split is a view of input.

If input is one dimensional this is equivalent to calling oneflow.tensor_split(input, indices_or_sections, dim=0) (the split dimension is zero), and if input has two or more dimensions it’s equivalent to calling oneflow.tensor_split(input, indices_or_sections, dim=1) (the split dimension is 1), except that if indices_or_sections is an integer it must evenly divide the split dimension or a runtime error will be thrown.

Parameters
Returns

the output TensorTuple.

Return type

oneflow.TensorTuple

For example:

>>> import oneflow as flow

>>> input = flow.rand(3,4,5,6)
>>> output = flow.hsplit(input,(1,3))
>>> output[0].size()
oneflow.Size([3, 1, 5, 6])
>>> output[1].size()
oneflow.Size([3, 2, 5, 6])
>>> output[2].size()
oneflow.Size([3, 1, 5, 6])