Hi organizers,
Thanks for organizing this awesome challenge. It seems all the data processing part is written by TensorFlow as well as the baseline models. As PyTorch is also very popular in the academia, would you mind to also provide some examples for the PyTorch users?
Best,
Posted by: HDXY @ Oct. 7, 2020, 8:23 p.m.Hi!
Thank you for your interest in the challenge! The data processing part of this challenge is mainly based on the repo Meta-dataset of Google, which is very well designed and coded. We focused on implementing TF baselines using the Autograph feature as it is similar to the way PyTorch code is implemented. The challenge do allow PyTorch implementations and you can find a way to process the data coming from a data generator for PyTorch here :
https://github.com/google-research/meta-dataset/blob/master/Intro_to_Metadataset.ipynb
The fourth part in this notebook is the one you are looking for. Due to the limitation of our manpower, we welcome community efforts to share a PyTorch example code to us all.
Thank you for your understanding!