frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-10-21 05:34:18 +08:00
Code Issues Packages Projects Releases Wiki Activity

Enable alternative LayerNorm impl in FisherGan (#12178)

Browse Source 
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12178

Fisher GAN calls processor_util.add_mlp, which inject the layer norm through the
normalizer. We allow to use alternative impl for LayerNorn in the normalizer.

Reviewed By: Wakeupbuddy

Differential Revision: D9235528

fbshipit-source-id: 88c126c658102926613242ef84a481f6de1676ed
This commit is contained in:
Xiaolong Wang
2018-10-11 17:33:31 -07:00
committed by Facebook Github Bot
parent 8ac8b823c2
commit 93a4b76114
1 changed files with 3 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
caffe2/python/normalizer.py
View File

@ -31,11 +31,12 @@ class BatchNormalizer(Normalizer):
class LayerNormalizer(Normalizer):
def __init__(self, epsilon):
def __init__(self, epsilon, use_layer_norm_op=True):
super(LayerNormalizer, self).__init__()
self._epsilon = float(epsilon)
self._use_layer_norm_op = use_layer_norm_op
def _run(self, layer_model, param):
return layer_model.LayerNormalization(
param, epsilon=self._epsilon
param, epsilon=self._epsilon, use_layer_norm_op=self._use_layer_norm_op
)