Yes. Unfortunately we can't trigger a weight update during forward computation automatically in MXNet as far as I am aware, so the user needs to trigger it manually if he wants to have it.
Alternatively to manually triggering it as introduced above, we can extend the `Trainer` and `Optimizer` APIs to support some kind of `apply_lazy_regularization` method that hides above logic and only requires the user to specify the set of rows to update. [ Full content available at: https://github.com/apache/incubator-mxnet/pull/12365 ] This message was relayed via gitbox.apache.org for [email protected]
