PyTorch Datasets

Series

PyTorch Datasets

Articles about data input pipelines for training neural networks in PyTorch.

Solving Core Data Loading Challenges with PyTorch
PyTorch data utilities define concepts of dataset and data loader. They share their general purpose of loading the training data, thus they are easily confused. Why do we need both of them? Building a deep learning input pipeline involves three chall...
Jul 5, 20209 min read131
FlatMapDataset: flat-map data transform for PyTorch
PyTorch established a simple a consistent toolset for building training pipelines. The most common scenarios can be handled with two utilities: the Dataset class and the DataLoader class. I wrote about them in my previous article. I've noticed some p...
Jul 22, 202014 min read296