search

isht7 / pytorch-deeplab-resnet

DeepLab resnet v2 model in pytorch
MIT License
602 stars 118 forks source link

model.eval() will change mean and var #23

Closed luhc15 closed 6 years ago

luhc15 commented 6 years ago

when set model.eval(), the BN layers weights and bias are fixed, but var and mean will changed when finetuned, is there any influences if var and mean change or should set them fixed using momentum=0?

omkar13 commented 6 years ago

When using model.eval(), mean and variance are fixed to pretrained values. BN layer's weights and bias can be fixed using requires_grad=False.

isht7 commented 6 years ago

In my model, I have set requres_grad = False for all batch-norm parameters. Thus weights of all BN layers remain fixed. model.eval() is also used, which keeps the running mean and var fixed.