diff options
| author |  Jianwei Xie <xiejw@google.com> | 2017-06-26 11:44:20 -0700 |
|---|---|---|
| committer |  TensorFlower Gardener <gardener@tensorflow.org> | 2017-06-26 11:48:37 -0700 |
| commit | ee3eaffe162da95e98d3d5ee8e42e296d8b66d1c (patch) | |
| tree | 8530ab0c6eefd50ba3a0ebf615fb5397f3b6ce08 | |
| parent | 9b2984d985357a02b39baa495b8d7c0fe9475377 (diff) | |
Adds notes to prevent overfitting for Experiment continous_train_and_eval.
PiperOrigin-RevId: 160172692
| -rw-r--r-- | tensorflow/contrib/learn/python/learn/experiment.py | 7 |
1 files changed, 6 insertions, 1 deletions
diff --git a/tensorflow/contrib/learn/python/learn/experiment.py b/tensorflow/contrib/learn/python/learn/experiment.py index c60ecac5df..491c4343bd 100644 --- a/tensorflow/contrib/learn/python/learn/experiment.py +++ b/tensorflow/contrib/learn/python/learn/experiment.py @@ -523,7 +523,12 @@ class Experiment(object): differences in resource control. First, the resources (e.g., memory) used by training will be released before evaluation (`train_and_evaluate` takes double resources). Second, more checkpoints will be saved as a checkpoint - is generated at the end of each small trainning iteration. + is generated at the end of each trainning iteration. + + 3. As the estimator.train starts from scratch (new graph, new states for + input, etc) at each iteration, it is recommended to have the + `train_steps_per_iteration` larger. It is also recommended to shuffle your + input. Args: continuous_eval_predicate_fn: A predicate function determining whether to |