orbit.StandardEvaluatorOptions

Advanced options for the orbit.StandardEvaluator.

use_tf_function A boolean indicating whether to apply tf.function to the evaluation loop. This will only affect the body of the loop (involving eval_step); eval_loop_begin and eval_loop_end will always be run in eager mode.
use_tf_while_loop A boolean indicating whether to run the evaluation loop using a tf.while_loop. If True, use_tf_function must also be True.
recreate_iterator_for_each_eval A boolean indicating whether to recreate a new iterator for the evaluation dataset before each round of evaluation, which implies each round of evaluation starts from the beginning of the evaluation dataset. For example, the evaluation dataset is [1, 2, 3, 4], batch size is 1 and evaluation steps is 2. If True, the data to be evaluated is [1, 2] every time. If False, the iterator state is maintained between calls to StandardEvaluator.evaluate().

Methods

__eq__

recreate_iterator_for_each_eval True
use_tf_function True
use_tf_while_loop False