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.
I don't see omniglot included in the github repository. There is no folder "datasets" (btw, I noticed that .gitignore is set to ignore datasets/*)
While the first link to download mini-imagenet is working, the second link (under "To run an experiment from the paper on Mini-Imagenet:") is broken. Also, this section refers to a .pbzip file, did you mean .tar.bz2?
Btw, thanks for making the code available. It is particularly nice that a "scripts" folder is included to make it easy to replicate the experiments.
Thanks for bringing this up to my attention. I deleted the omniglot dataset by mistake. I've now added it again. Also, I fixed the mini-imagenet links. :)
I have two comments about the datasets:
Btw, thanks for making the code available. It is particularly nice that a "scripts" folder is included to make it easy to replicate the experiments.