Batchnorm.eval() cause worst result

[jabacrack]

model
a = [Sequential(convolution(), batchnorm(),activation_f())] * n
b = [Sequential(Upsample(), batchnorm(), convolution(), batchnorm(), activation_f(), convolution(), batchnorm(), activation_f())] * n
model = Sequential(*a, *b)

@ptrblck

Have you changed the momentum in BatchNorm?

no

@ptrblck

Are you preprocessing the data differently?

no

@SimonW

Also, what was your training batch size?

I try with only one batch