torch.sparse.softmax

torch.sparse.softmax(input, dim, dtype=None)

Applies a softmax function.

Softmax is defined as:

$\text{Softmax}(x_{i}) = \frac{exp(x_i)}{\sum_j exp(x_j)}$

where $i, j$ run over sparse tensor indices and unspecified entries are ignores. This is equivalent to defining unspecified entries as negative infinity so that $exp(x_k) = 0$ when the entry with index $k$ has not specified.

It is applied to all slices along dim, and will re-scale them so that the elements lie in the range [0, 1] and sum to 1.

Parameters

• input (Tensor) – input

• dim (int) – A dimension along which softmax will be computed.

• dtype (torch.dtype, optional) – the desired data type of returned tensor. If specified, the input tensor is casted to dtype before the operation is performed. This is useful for preventing data type overflows. Default: None