diff options
| author |  Sergio Guadarrama <sguada@google.com> | 2016-07-19 10:27:52 -0800 |
|---|---|---|
| committer |  TensorFlower Gardener <gardener@tensorflow.org> | 2016-07-19 11:32:52 -0700 |
| commit | 580cdddab964d65e45d9bfeb7b9eca43f18db7eb (patch) | |
| tree | 677bf107cb2821b273fa150d8ab243773107bc7c | |
| parent | 4307ebc0d79b7f725c4b309bff5d73a7506ac720 (diff) | |
Updated comment about batch_norm layer.
Change: 127853412
| -rw-r--r-- | tensorflow/contrib/layers/python/layers/layers.py | 5 |
1 files changed, 4 insertions, 1 deletions
diff --git a/tensorflow/contrib/layers/python/layers/layers.py b/tensorflow/contrib/layers/python/layers/layers.py index 7588daa8f4..e4a25fa113 100644 --- a/tensorflow/contrib/layers/python/layers/layers.py +++ b/tensorflow/contrib/layers/python/layers/layers.py @@ -215,10 +215,13 @@ def batch_norm(inputs, trainable=False, collections=moving_variance_collections) + # If `is_training` doesn't have a constant value, because it is a `Tensor`, + # a `Variable` or `Placeholder` then is_training_value will be None and + # `needs_moments` will be true. is_training_value = utils.constant_value(is_training) - # Calculate the moments based on the individual batch. need_moments = is_training_value is None or is_training_value if need_moments: + # Calculate the moments based on the individual batch. mean, variance = nn.moments(inputs, axis, shift=moving_mean) moving_vars_fn = lambda: (moving_mean, moving_variance) if updates_collections is None: |