The original code for the paper "How to train your MAML" along with a replication of the original "Model Agnostic Meta Learning" (MAML) paper in Pytorch.
May I ask how to apply the dataparallel function of Pytorch on the network in your implementation? Faster training and larger batchsize can be achieved if parallelizing the MAML computational graph with dataparallel function. This currently seems not easy to do with Pytorch. I am very appreciated if you have any idea on this point.
May I ask how to apply the dataparallel function of Pytorch on the network in your implementation? Faster training and larger batchsize can be achieved if parallelizing the MAML computational graph with dataparallel function. This currently seems not easy to do with Pytorch. I am very appreciated if you have any idea on this point.