The HuggingFace ecosystem consists of mainly two parts: a family of libraries and the Hub. The libraries provide the code while the Hub provides the pretrained model weights, datasets, scripts for the evaluation metrics, and more.
https://www.youtube.com/watch?v=3kRB2TXewus
As outlined earlier, transfer learning is one of the key factors driving the success of transformers because it makes it possible to reuse pretrained models for new tasks. Consequently, it is crucial to be able to load pretrained models quickly and run experiments with them.
The Hugging Face Hub hosts over 20,000 freely available models. As shown in Figure below, there are filters for tasks, frameworks, datasets, and more that are designed to help you navigate the Hub and quickly find promising candidates. As we've seen with the pipelines, loading a promising model in your code is then literally just one line of code away. This makes experimenting with a wide range of models simple, and allows you to focus on the domain-specific parts of your project.
In addition to model weights, the Hub also hosts datasets and scripts for computing metrics, which let you reproduce published results or leverage additional data for your application.