giagrad.Tensor.log_softmax

Tensor.log_softmax(axis: int) → Tensor

Applies LogSoftmax function to every 1-D slice defined by axis.

LogSoftmax for a one-dimensional slice is defined as:

\[\text{LogSoftmax}(x_i) = \log \left( \frac{\exp(x_i)}{\sum_j \exp(x_j)} \right)\]

Parameters:

axis (int) – The dimension along which LogSoftmax will be computed.

Examples

>>> t = Tensor.empty(2, 3).uniform(-1, 1)
>>> t
tensor: [[-0.07469178  0.7226724   0.98966014]
         [-0.01990889 -0.4521888   0.26520386]]
>>> t.softmax(axis=1)
tensor: [[-0.72091377 -0.26915795 -0.39513725]
         [-0.6661309  -1.4440191  -1.1195936 ]] fn: LogSoftmax(axis=0)