[UNet Industrial for TensorFlow] the occupation of gpu memory during training

[hum-yong]

Related to "UNet Industrial for TensorFlow"

Describe when I train this model with a custom dataset(5000 files) , my computer will get stuck. when 3150 files are used, the occupation of gpu memory has reached 8GB. I tried to modify some training parameters, but it didn't work. who can provide some methods to reduce gpu memory consumption during training

Environment Container : TensorFlow 21.03-tf1-py3 GPU: one rtx3060-12GB