What is the meaning of “config.FIXED_PARAMS” in config.py?


#1

The path is “incubator-mxnet/example/rcnn/rcnn/config.py”.

I want to know the “config.FIXED_PARAMS = [‘conv1’, ‘conv2’]”, "config.FIXED_PARAMS_SHARED ", “network.resnet.FIXED_PARAMS”, "network.resnet.FIXED_PARAMS_SHARED ".

Thank you.


#2

I am not super familiar with that example but looking through the code it seems to be a prefix matching for layers whose weights are frozen at training time, i.e the weights are not updated during training.


#4

Thank you!
I want to know how to take control of every layer update or not update.
Dose use “network.resnet.FIXED_PARAMS = [‘conv0’, ‘stage1’, ‘gamma’, ‘beta’]
network.resnet.FIXED_PARAMS_SHARED = [‘conv0’, ‘stage1’, ‘stage2’, ‘stage3’, ‘gamma’, ‘beta’]” do it ?