hypnettorch - Hypernetworks in PyTorch

This package provides functionalities to easily work with hypernetworks in PyTorch. A hypernetwork h(\mathbf{e}, \theta) is a neural network with parameters \theta that generates the parameters \omega of another neural network f(\mathbf{x}, \omega), called main network. These two network types require specialized implementations. For instance, a main network must have the ability to receive its own weights \omega as additional input to the forward method (see subpackage mnets). A collection of different hypernetwork implementations can be found in subpackage hnets.


See here.


Check out the tutorials, especially the getting started tutorial.

You can also check out example implementations that make use of hypnettorch.

Indices and tables