Getting started

Working with FFCV just requires two steps:

1. Converting an existing dataset into FFCV format, and

2. Making FFCV data loaders.

In this section, we’ll go over these steps one by one, and go from scratch to a complete data loader ready for training.

• Writing a dataset to FFCV format
  • Indexable Dataset
  • Webdataset
  • Fields
• Making an FFCV dataloader
  • Dataset ordering
  • Pipelines
  • Putting together
  • Other options
  • More information