Biblioteca Abierta · SciTechBiz Institute

The Hugging Face Course

Hugging Face course on Transformer models and language models: using the Transformers library, fine-tuning, Datasets and Tokenizers, classical NLP tasks, Gradio demos, data curation, LLM fine-tuning and reasoning models. SciTechBiz Institute training based on open materials from Hugging Face, under the Apache 2.0 licence.

  • Hugging Face · Apache 2.0
  • Lecciones: 81
  • ≈ 34 h
  • inglés
También disponible en castellano

Lo que aprenderás

Programa

SetupLecciones: 1
  1. Introduction
Transformer modelsLecciones: 9
  1. Introduction
  2. Natural Language Processing and Large Language Models
  3. Transformers, what can they do?
  4. How do Transformers work?
  5. How Transformers solve tasks
  6. Transformer Architectures
  7. Deep dive into Text Generation Inference with LLMs
  8. Bias and limitations
  9. Summary
Using TransformersLecciones: 8
  1. Introduction
  2. Behind the pipeline
  3. Models
  4. Tokenizers
  5. Handling multiple sequences
  6. Putting it all together
  7. Basic usage completed!
  8. Optimized Inference Deployment
Fine-tuning a pretrained modelLecciones: 6
  1. Introduction
  2. Processing the data
  3. Fine-tuning a model with the Trainer API
  4. A full training loop
  5. Understanding Learning Curves
  6. Fine-tuning, Check!
Sharing models and tokenizersLecciones: 5
  1. The Hugging Face Hub
  2. Using pretrained models
  3. Sharing pretrained models
  4. Building a model card
  5. Part 1 completed!
The Datasets libraryLecciones: 7
  1. Introduction
  2. What if my dataset isn't on the Hub?
  3. Time to slice and dice
  4. Big data? Datasets to the rescue!
  5. Creating your own dataset
  6. Semantic search with FAISS
  7. Datasets, check!
The Tokenizers libraryLecciones: 10
  1. Introduction
  2. Training a new tokenizer from an old one
  3. Fast tokenizers' special powers
  4. Fast tokenizers in the QA pipeline
  5. Normalization and pre-tokenization
  6. Byte-Pair Encoding tokenization
  7. WordPiece tokenization
  8. Unigram tokenization
  9. Building a tokenizer, block by block
  10. Tokenizers, check!
Classical NLP tasksLecciones: 8
  1. Introduction
  2. Token classification
  3. Fine-tuning a masked language model
  4. Translation
  5. Summarization
  6. Training a causal language model from scratch
  7. Question answering
  8. Mastering LLMs
Building and sharing demosLecciones: 8
  1. Introduction to Gradio
  2. Building your first demo
  3. Understanding the Interface class
  4. Sharing demos with others
  5. Integrations with the Hugging Face Hub
  6. Advanced Interface features
  7. Introduction to Gradio Blocks
  8. Gradio, check!
Curate high-quality datasetsLecciones: 6
  1. Introduction to Argilla
  2. Set up your Argilla instance
  3. Load your dataset to Argilla
  4. Annotate your dataset
  5. Use your annotated dataset
  6. Argilla, check!
Fine-tune Large Language ModelsLecciones: 6
  1. Supervised Fine-Tuning
  2. Chat Templates
  3. Supervised Fine-Tuning
  4. LoRA (Low-Rank Adaptation)
  5. Evaluation
  6. Conclusion
Build Reasoning ModelsLecciones: 7
  1. Open R1 for Students
  2. Introduction to Reinforcement Learning and its Role in LLMs
  3. Understanding the DeepSeek R1 Paper
  4. Advanced Understanding of Group Relative Policy Optimization (GRPO) in DeepSeekMath
  5. Implementing GRPO in TRL
  6. Practical Exercise: Fine-tune a model with GRPO
  7. Practical Exercise: GRPO with Unsloth

Certificado y tutor de IA

Los miembros estudian estos cursos sin coste adicional. El certificado de aprovechamiento del SciTechBiz Institute y el tutor de IA se incluyen con la licencia.

Ver la licencia

Atribución

Fuente
Hugging Face
Licencia
Apache 2.0
Versión
8b893f0ede6c

Based on "The Hugging Face Course" by Hugging Face (GitHub, Apache 2.0 licence)

Hugging Face does not endorse or certify this training.

Las fuentes citadas no respaldan ni certifican estos cursos; se usan bajo sus licencias abiertas, con atribución en cada lección.