RETHINKING DATA LOADING IN PYTORCH | VITALY FEDYUNIN
Data is one of the most important aspects of machine learning research, something the PyTorch team is always actively looking to improve through data-related utilities. In this talk, Vitaly Fedyunin (Software Engineer, Meta AI) discusses how to think about data preparation, how to decompose datasets, and setting up dataloader to get the most of your data. He also gives you a peek into ways the PyTorch team is looking ahead improving existing data loader and support with experimental features.