Do you need to construct the network twice for training and inference accordingly?

jianlong-yuan / syncbn-tensorflow

Synchronized Multi-GPU Batch Normalization

67 stars 13 forks source link

Do you need to construct the network twice for training and inference accordingly? #2

Closed panfengli closed 3 years ago

panfengli commented 5 years ago

Since tf.nn.fused_batch_norm will need a python bool option, which is different from tf.layers.batch_normalization where a placeholder can be used for training argument.

jianlong-yuan commented 5 years ago

You can just use python bool op