StudyVisuallyStart studying
AI developers · Part 07

Hugging Face

Open models, the Hub, pipelines and hosted inference.

Last updated

  1. The Hugging Face HubA hub with hundreds of thousands of free, open-source models, plus libraries to run them.
  2. Transformers.jsA library (@huggingface/transformers) that runs models directly in Node or the browser. No Python.
  3. pipeline()One function that loads a model for a task (embeddings, classification, and more) and gives you a callable.
  4. Local embeddingsRun an embedding model like all-MiniLM-L6-v2 on your own server. It downloads once (tens of MB), then runs on CPU.
  5. Zero-shot classificationLabel text with categories you choose on the spot, without training anything.
  6. Inference APICall hosted Hugging Face models over HTTP with an HF token (@huggingface/inference).
  7. Self-hosted modelsRun open LLMs like Llama or Mistral on your own server with tools like Ollama or TGI.
  8. Model cardThe README of a model: what it does, its license, and its limits.
  9. Open weightsYou can download the model itself and run it yourself (Llama, Mistral, Qwen).
  10. Fine-tuningTraining an existing model further on your own examples to change its behaviour or style.
  11. QuantizationShrinking a model (for example to 4-bit numbers) so it runs on smaller hardware, with a small quality loss.
  12. OpenAI vs Hugging FaceUse OpenAI (or Claude) for reasoning and tool calling. Use small Hugging Face models for high-volume embeddings, classification and private data.