Why it matters

Hub is where you find models to use and where you share your own. Central to modern ML practice.

Advertisement

The architecture

Repository structure: git repo with model weights (safetensors, pytorch_model.bin), config, tokenizer files, README.md as model card.

Hub UI: browse, search, filter by task/library/license.

HuggingFace Hub contentsModelsweights + configDatasetscurated dataSpacesinteractive demosEvery repo is a Git repo; version by commit hash or tag
Hub content types.
Advertisement

How it works end to end

Loading: from_pretrained('user/model') fetches from Hub, caches under ~/.cache/huggingface.

Private repos: authentication token, gated access, or paid tier.

Model cards: README.md documents intended use, training data, limitations, license.

Spaces: Gradio/Streamlit apps hosted for demos.