I saw in [this](https://github.com/apache/incubator-mxnet/issues/1340#issuecomment-174166248) and [this](https://github.com/apache/incubator-mxnet/issues/3073#issuecomment-241033513) issue comments, @winstywang suggests that using BlockGrad with weight decay is highly inadviseable. As I also asked in [this](https://github.com/apache/incubator-mxnet/issues/3073#issuecomment-416852189) comment, how can I use BlockGrad to freeze first layers of the net, while still using weight decay on the unfrozen ones?
[ Full content available at: https://github.com/apache/incubator-mxnet/issues/12392 ] This message was relayed via gitbox.apache.org for [email protected]
