I have used the resnet-binary model in BMXnet to train my own data, but get float weights not binary weights. Why?
The parameter of the network
Load the trained model and use model.get_params()
I have used the resnet-binary model in BMXnet to train my own data, but get float weights not binary weights. Why?
The parameter of the network
Load the trained model and use model.get_params()
It looks like there are steps necessary to binarize the model. See that project's repo: https://github.com/hpi-xnor/BMXNet
You may get more specialized assistance by posting in the issues tab of that repo.