mxnet-the-straight-dope
mxnet-the-straight-dope copied to clipboard
`is_training` for Batch Normalization in gluon
In Batch Normalization from scratch, there is a is_training
parameter to control the behavior of Batch Normalization. However, for Batch Normalization in gluon, there is no such an option. How do we know it's using global moving statistics or local batch-norm in GLUON?
Same doubt here! Could you please clarify? Thank you, Patricia
@patriciaaa82 @jonbakerfish I think you may find the answer here https://gluon.mxnet.io/chapter03_deep-neural-networks/mlp-dropout-gluon.html#Accessing-is_training()-status