=================================================================================

Both tf.contrib.estimator.stop_if_no_decrease_hook and tf.estimator.experimental.stop_if_no_decrease_hook are used for early stopping for the training based on the evaluation loss. The early-stopping hook uses the evaluation results to decide when it's time to cut the training, but you need to pass in the number of training steps that you want to monitor and keep in mind how many evaluations will happen in that number of steps. For instance, if you set the hook as hook = early_stopping.stop_if_no_decrease_hook(estimator, "loss", 10000) the hook will consider the evaluations happening in a range of 10k steps.

============================================