torch.nonzero (original) (raw)
Note
torch.nonzero(..., as_tuple=False) (default) returns a 2-D tensor where each row is the index for a nonzero value.
torch.nonzero(..., as_tuple=True) returns a tuple of 1-D index tensors, allowing for advanced indexing, so x[x.nonzero(as_tuple=True)]
gives all nonzero values of tensor x
. Of the returned tuple, each index tensor contains nonzero indices for a certain dimension.
See below for more details on the two behaviors.
When input
is on CUDA, torch.nonzero() causes host-device synchronization.
When as_tuple
is False
(default):
Returns a tensor containing the indices of all non-zero elements ofinput
. Each row in the result contains the indices of a non-zero element in input
. The result is sorted lexicographically, with the last index changing the fastest (C-style).
If input
has nn dimensions, then the resulting indices tensorout
is of size (z×n)(z \times n), where zz is the total number of non-zero elements in the input
tensor.
When as_tuple
is True
:
Returns a tuple of 1-D tensors, one for each dimension in input
, each containing the indices (in that dimension) of all non-zero elements ofinput
.
If input
has nn dimensions, then the resulting tuple contains nntensors of size zz, where zz is the total number of non-zero elements in the input
tensor.
As a special case, when input
has zero dimensions and a nonzero scalar value, it is treated as a one-dimensional tensor with one element.
Parameters
input (Tensor) – the input tensor.
Keyword Arguments
out (LongTensor , optional) – the output tensor containing indices
Returns
If as_tuple
is False
, the output tensor containing indices. If as_tuple
is True
, one 1-D tensor for each dimension, containing the indices of each nonzero element along that dimension.
Return type
LongTensor or tuple of LongTensor
Example:
torch.nonzero(torch.tensor([1, 1, 1, 0, 1])) tensor([[ 0], [ 1], [ 2], [ 4]]) torch.nonzero(torch.tensor([[0.6, 0.0, 0.0, 0.0], ... [0.0, 0.4, 0.0, 0.0], ... [0.0, 0.0, 1.2, 0.0], ... [0.0, 0.0, 0.0,-0.4]])) tensor([[ 0, 0], [ 1, 1], [ 2, 2], [ 3, 3]]) torch.nonzero(torch.tensor([1, 1, 1, 0, 1]), as_tuple=True) (tensor([0, 1, 2, 4]),) torch.nonzero(torch.tensor([[0.6, 0.0, 0.0, 0.0], ... [0.0, 0.4, 0.0, 0.0], ... [0.0, 0.0, 1.2, 0.0], ... [0.0, 0.0, 0.0,-0.4]]), as_tuple=True) (tensor([0, 1, 2, 3]), tensor([0, 1, 2, 3])) torch.nonzero(torch.tensor(5), as_tuple=True) (tensor([0]),)