tf.compat.v2.debugging.assert_near

Assert the condition x and y are close element-wise.

This Op checks that x[i] - y[i] < atol + rtol * tf.abs(y[i]) holds for every pair of (possibly broadcast) elements of x and y. If both x and y are empty, this is trivially satisfied.

If any elements of x and y are not close, message, as well as the first summarize entries of x and y are printed, and InvalidArgumentError is raised.

The default atol and rtol is 10 * eps, where eps is the smallest representable positive number such that 1 + eps != 1. This is about 1.2e-6 in 32bit, 2.22e-15 in 64bit, and 0.00977 in 16bit. See numpy.finfo.

Args
x Float or complex Tensor.
y Float or complex Tensor, same dtype as and broadcastable to x.
rtol Tensor. Same dtype as, and broadcastable to, x. The relative tolerance. Default is 10 * eps.
atol Tensor. Same dtype as, and broadcastable to, x. The absolute tolerance. Default is 10 * eps.
message A string to prefix to the default message.
summarize Print this many entries of each tensor.
name A name for this operation (optional). Defaults to "assert_near".
Returns
Op that raises InvalidArgumentError if x and y are not close enough. This can be used with tf.control_dependencies inside of tf.functions to block followup computation until the check has executed.
Raises
InvalidArgumentError if the check can be performed immediately and x != y is False for any pair of elements in x and y. The check can be performed immediately during eager execution or if x and y are statically known.

Eager Compatibility

returns None

Numpy Compatibility

Similar to numpy.assert_allclose, except tolerance depends on data type. This is due to the fact that TensorFlow is often used with 32bit, 64bit, and even 16bit data.

© 2020 The TensorFlow Authors. All rights reserved.
Licensed under the Creative Commons Attribution License 3.0.
Code samples licensed under the Apache 2.0 License.
https://www.tensorflow.org/versions/r1.15/api_docs/python/tf/compat/v2/debugging/assert_near