Missed TensorFlow World? Check out the recap. Learn more

tf.test.compute_gradient

TensorFlow 1 version View source on GitHub

Computes the theoretical and numeric Jacobian of f.

Aliases:

tf.test.compute_gradient(
    f,
    x,
    delta=0.001
)

With y = f(x), computes the theoretical and numeric Jacobian dy/dx.

Args:

  • f: the function.
  • x: a list arguments for the function
  • delta: (optional) perturbation used to compute numeric Jacobian.

Returns:

A pair of lists, where the first is a list of 2-d numpy arrays representing the theoretical Jacobians for each argument, and the second list is the numerical ones. Each 2-d array has "x_size" rows and "y_size" columns where "x_size" is the number of elements in the corresponding argument and "y_size" is the number of elements in f(x).

Raises:

  • ValueError: If result is empty but the gradient is nonzero.
  • ValueError: If x is not list, but any other type.

Example:

@tf.function
def test_func(x):
  return x*x

theoretical, numerical = tf.test.compute_gradient(test_func, [1.0])
theoretical, numerical
# ((array([[2.]], dtype=float32),), (array([[2.000004]], dtype=float32),))