Hugging Face & Local LLM Tutorials

Practical guides for working with Hugging Face Transformers and running large language models locally.

Models & storage

• How to download Hugging Face models to a custom cache folder (Python) — cache_dir, HF_HOME, TRANSFORMERS_CACHE, and save_pretrained.

Run LLMs locally

• Convert a Hugging Face model to GGUF in Google Colab — quantize and export for llama.cpp / Ollama.
• Run FLAN-T5 and GPT large language models locally with Gradio — a local web UI in Python.

Build models from scratch

• Build a basic LLM (GPT) from scratch in Python
• Text generation from scratch: build a generative AI model in Python

Related topics

Continue with Generative AI, AI Agents, and the daily AI Rankings of top AI repositories, papers, and packages.

FAQ

How do I stop Hugging Face from filling up my default drive? Point the cache elsewhere with cache_dir / HF_HOME — see downloading models to a custom cache folder.

How do I run a Hugging Face model offline / locally? Convert it to GGUF in Google Colab, then run it locally, or use a local Gradio app for FLAN-T5 / GPT.