This variant on the TorchMeta MAML example shows how higher can work well with TorchMeta to allow MAML implementations as minimal changes to a pure pytorch codebase. TorchMeta is used exclusively for data loaders/utils, while higher deals with the unroll, allowing users to use arbitrary models and optimizers written in pure pytorch.
This is just a suggested addition to the codebase to show interoperability between the libraries, but if you feel it doesn't "belong" in the codebase, I certainly won't feel offended 😉
This variant on the TorchMeta MAML example shows how higher can work well with TorchMeta to allow MAML implementations as minimal changes to a pure pytorch codebase. TorchMeta is used exclusively for data loaders/utils, while higher deals with the unroll, allowing users to use arbitrary models and optimizers written in pure pytorch.
This is just a suggested addition to the codebase to show interoperability between the libraries, but if you feel it doesn't "belong" in the codebase, I certainly won't feel offended 😉