sofmax函数 tf.nn.softmax(logits, dim=-1, name=None)

softmax(logits, dim=-1, name=None):
  """Computes softmax activations.

  For each batch `i` and class `j` we have

                                softmax = exp(logits) / reduce_sum(exp(logits), dim)

  Args:
    logits:  A non-empty `Tensor`. Must be one of the following types: `half`,  `float32`, `float64`.
    dim: The dimension softmax would be performed on. The default is -1 which indicates the last dimension.
    name: A name for the operation (optional).

  Returns:
    A `Tensor`. Has the same type as `logits`. Same shape as `logits`.
  Raises:
    Invalid Argument Error: if `logits` is empty or `dim` is beyond the last dimension of `logits`.