₹0 for students · Certificate · Intermediate

The Hugging Face LLM Course

14 lessons: use, fine-tune and share transformer models with the Hugging Face ecosystem

Start free in the app

14 lessons · ~11 hours · 28-question final exam · pass mark 70%

About the course

What you’ll learn

Learn the open-source tools that most AI teams use: Transformers pipelines, models and tokenizers; fine-tuning with the Trainer API; the Hub, model cards and Datasets; how tokenizers work; classic NLP tasks from named-entity recognition to question answering; debugging; Gradio demos; curating data with Argilla; fine-tuning LLMs with SFT and LoRA; and building reasoning models with GRPO. Each lesson has a study guide, links to the interactive chapter and notebooks, and a quiz. Pass every quiz and the final exam to earn a verifiable certificate.

Hugging Face TransformersTokenizersFine-tuningDatasetsGradioLoRASFTGRPONLP tasksModel cards

Part of the career track: GenAI and LLM Builder

How the certificate works

  1. Read each lesson’s study guide, then the full lesson and any notebooks or exercises.
  2. Pass every lesson quiz with 70% or more.
  3. Pass the final exam: 28 questions in 40 minutes.
  4. Get a certificate with a public verification link, and add it to LinkedIn.

Syllabus

14 lessons in 3 parts

Transformers and the Hugging Face ecosystem

1
Transformer Models and What They Can Do
45 min · NLP tasks: classification, NER, QA, summarisation, translation, generation.
2
Inference with LLMs, Bias and Limitations
35 min · Inference: prefill processes the prompt; decode generates tokens one by one.
3
Using Transformers: Models and Tokenizers
50 min · Pipeline = tokenizer → model (logits) → postprocessing.
4
Fine-Tuning a Pretrained Model
55 min · Fine-tune pretrained models on task data, e.g. BERT on MRPC paraphrases.
5
The Hub: Sharing Models and Model Cards
35 min · The Hub hosts models, datasets and Spaces as Git repositories with history.

Data, tokenizers and NLP tasks

6
The Datasets Library
50 min · load_dataset reads Hub datasets and local/remote files (CSV, JSON, Parquet).
7
The Tokenizers Library
50 min · Train a new tokenizer for your domain or language with train_new_from_iterator.
8
NLP Tasks: Token Classification and Masked Language Modelling
50 min · Token classification labels each token: NER, POS tagging, chunking.
9
NLP Tasks: Translation, Summarisation, Language Models and QA
60 min · Translation: fine-tune Marian with Seq2SeqTrainer; evaluate with BLEU.

Shipping and advanced LLMs

10
Debugging and Asking for Help
35 min · Read tracebacks from the bottom; search exact error messages.
11
Building and Sharing Demos with Gradio
40 min · Demos make models usable by anyone; Gradio wraps Python functions with web UIs.
12
Curating High-Quality Datasets with Argilla
35 min · Data quality drives model quality; Argilla supports human annotation.
13
Fine-Tuning LLMs: Chat Templates, SFT and LoRA
55 min · Chat templates format role-tagged messages into the tokens a model expects; use apply_chat_template.
14
Building Reasoning Models with Reinforcement Learning
50 min · Reasoning models think step by step; RL trains them with rewards rather than imitation.

This course is based on the Hugging Face LLM Course (https://huggingface.co/learn/llm-course; https://github.com/huggingface/course), licensed under the Apache License 2.0. Vikosha wrote new study guides and quiz questions; changes were made. Certificates are issued by Vikosha and are not issued or endorsed by Hugging Face. Licence: Apache License 2.0, Copyright Hugging Face. View the original curriculum