Player not loading? Watch on YouTube
This tutorial covers Hugging Face CLI workflows for downloading models and managing Hub resources from a terminal. The presenter introduces installation commands for Mac and Windows, then demonstrates browser login with an authorization code. Model searches include a BERT filter, a result limit and JSON output piped to jq.
For local AI workflows, the download example uses an include filter to select a particular quantization and a custom destination directory. A dry run previews the files and their sizes before downloading. The cache section shows how to list entries, sort them by size, remove an unused model and prune incomplete or unconnected downloads. The tutorial covers model acquisition and storage rather than inference.
The presenter also creates a model repository, changes its visibility to private and uploads a README from the terminal. A static Hugging Face Space demonstrates deployment through Git: clone the repository, edit its HTML, commit and push. These examples use Hugging Face hosting rather than a self-hosted service.
Later examples sync local files to a Bucket and query a recipe dataset through a Parquet URL. SQL results are limited to five rows, formatted as JSON and filtered with jq. The final section sorts papers by trending status and saves a paper as Markdown for preview in an editor.
