Introduction · Hugging Face

Hugging Face LLM Course Overview

• Course Objective: A free, ad-free curriculum teaching Large Language Models (LLMs) and Natural Language Processing (NLP) using the Hugging Face ecosystem (Transformers, Datasets, Tokenizers, Accelerate, and the Hub).
• Prerequisites: Strong Python knowledge and familiarity with introductory deep learning (e.g., fast.ai or DeepLearning.AI). Prior PyTorch or TensorFlow knowledge is helpful but not required.
• Structure:
  • Chapters 1–4: Introduction to the Transformers library, model usage, fine-tuning, and sharing on the Hub.
  • Chapters 5–8: Basics of Datasets and Tokenizers, classic NLP tasks, and LLM techniques.
  • Chapter 9: Building and sharing model demos via the Hub.
  • Chapters 10–12: Advanced topics including fine-tuning, high-quality dataset curation, and reasoning models.
• Time Commitment: Designed for 1 week per chapter, requiring approximately 6–8 hours of study per week.
• Resources:
  • Code is available via Google Colab, Amazon SageMaker Studio Lab, or the huggingface/notebooks GitHub repository.
  • Questions can be directed to the Hugging Face forums via the "Ask a question" banner.
• Certification: No official certification is currently available, though a program is in development.
• Licensing: The course is open-source under the Apache 2 license.
• Accessibility: The course is translated into multiple languages by the community, with ongoing work for additional language support.
• Authors: Developed by a team of machine learning engineers and researchers at Hugging Face, including Abubakar Abid, Ben Burtenshaw, Matthew Carrigan, Lysandre Debut, Sylvain Gugger, Dawood Khan, Merve Noyan, Lucile Saulnier, Lewis Tunstall, and Leandro von Werra.