torch.unique

torch.unique(input, sorted=True, return_inverse=False, return_counts=False, dim=None) → Tuple[Tensor, Tensor, Tensor]

Returns the unique elements of the input tensor.

Note

This function is different from torch.unique_consecutive() in the sense that this function also eliminates non-consecutive duplicate values.

Note

Currently in the CUDA implementation and the CPU implementation when dim is specified, torch.unique always sort the tensor at the beginning regardless of the sort argument. Sorting could be slow, so if your input tensor is already sorted, it is recommended to use torch.unique_consecutive() which avoids the sorting.

Parameters

• input (Tensor) – the input tensor

• sorted (bool) – Whether to sort the unique elements in ascending order before returning as output.

• return_inverse (bool) – Whether to also return the indices for where elements in the original input ended up in the returned unique list.

• return_counts (bool) – Whether to also return the counts for each unique element.

• dim (int) – the dimension to apply unique. If None, the unique of the flattened input is returned. default: None

Returns

A tensor or a tuple of tensors containing

• output (Tensor): the output list of unique scalar elements.

• inverse_indices (Tensor): (optional) if return_inverse is True, there will be an additional returned tensor (same shape as input) representing the indices for where elements in the original input map to in the output; otherwise, this function will only return a single tensor.

• counts (Tensor): (optional) if return_counts is True, there will be an additional returned tensor (same shape as output or output.size(dim), if dim was specified) representing the number of occurrences for each unique value or tensor.

Return type

(Tensor, Tensor (optional), Tensor (optional))

Example:

>>> output = torch.unique(torch.tensor([1, 3, 2, 3], dtype=torch.long))
>>> output
tensor([ 2,  3,  1])

>>> output, inverse_indices = torch.unique(
...     torch.tensor([1, 3, 2, 3], dtype=torch.long), sorted=True, return_inverse=True)
>>> output
tensor([ 1,  2,  3])
>>> inverse_indices
tensor([ 0,  2,  1,  2])

>>> output, inverse_indices = torch.unique(
...     torch.tensor([[1, 3], [2, 3]], dtype=torch.long), sorted=True, return_inverse=True)
>>> output
tensor([ 1,  2,  3])
>>> inverse_indices
tensor([[ 0,  2],
        [ 1,  2]])