cambridge-mlg / cnaps

Code for: "Fast and Flexible Multi-Task Classification Using Conditional Neural Adaptive Processes" and "TaskNorm: Rethinking Batch Normalization for Meta-Learning"
MIT License
159 stars 22 forks source link

from meta_dataset.data import * #7

Closed liulu112601 closed 4 years ago

liulu112601 commented 4 years ago

Hi, Thanks for the code! one quick question: https://github.com/cambridge-mlg/cnaps/blob/master/src/meta_dataset_reader.py#L8 from meta_dataset.data import dataset_spec as dataset_spec_lib

is the meta_dataset.data comes from the original meta-dataset github repo?

jfb54 commented 4 years ago

Yes. We use the episode reader from the meta-dataset repository. Step 2 in our installation instructions says:

Configure Meta-Dataset:

Also, before you run CNAPs, you need to export the location of the meta-dataset source code (see Step 1 in our usage section): export META_DATASET_ROOT=<root directory of the cloned or downloaded Meta-Dataset repository>