Missed TensorFlow Dev Summit? Check out the video playlist. Watch recordings


TensorFlow 1 version

Computes exp(x) - 1 element-wise.

    x, name=None

i.e. exp(x) - 1 or e^(x) - 1, where x is the input tensor. e denotes Euler's number and is approximately equal to 2.718281.

x = tf.constant(2.0)
tf.math.expm1(x) ==> 6.389056

x = tf.constant([2.0, 8.0])
tf.math.expm1(x) ==> array([6.389056, 2979.958], dtype=float32)

x = tf.constant(1 + 1j)
tf.math.expm1(x) ==> (0.46869393991588515+2.2873552871788423j)


  • x: A Tensor. Must be one of the following types: bfloat16, half, float32, float64, complex64, complex128.
  • name: A name for the operation (optional).


A Tensor. Has the same type as x.