Replace efficientnetv2 normalization layer

[Javier]

Hello,

How can I replace the 'normalization' layer in order to update the imagenet mean and variance values with my own values to fine-tune with my dataset.

I tried this, but the trained model has the original imagenet values

model = NetB0(weights="imagenet", include_top=False)

norm_index = model.layers.index(model.get_layer('normalization'))
model.layers[norm_index] = keras.layers.Normalization(mean=new_mean, variance=new_variance)
model.compile...

model.fit...

Thanks