The Hugging Face Course: Transformers and NLP

About this course

The official video lessons of the Hugging Face course: how transformer models work and how to use, fine-tune and share them with the Transformers, Datasets and Tokenizers libraries. Who it's for: developers and data scientists who want to work with modern NLP models. These are the free, official videos of the "Hugging Face Course" playlist from the Hugging Face YouTube channel (youtube.com/@huggingface). All videos belong to Hugging Face; Sikhami is not affiliated with Hugging Face and simply arranges the public playlist into a course, so you can learn one lesson at a time and keep track of your progress.

What you will learn

• The transformer architecture and what happens inside a pipeline • Tokenizers: word, character and subword tokenisation • Loading and processing data with the Datasets library • Fine-tuning a model with PyTorch or TensorFlow • Sharing models on the Model Hub • Tasks such as token classification, question answering and translation • Text embeddings, semantic search and debugging training

Course content

1 section · 79 lectures · 13h 32m

1. 1. Hugging Face Course

   • Welcome to the Hugging Face course4m
   • The pipeline function4m
   • The carbon footprint of Transformers5m
   • What is Transfer Learning?4m
   • The Transformer architecture2m
   • Transformer models: Encoders4m
   • Transformer models: Decoders4m
   • Transformer models: Encoder-Decoders6m
   • What happens inside the pipeline function? (PyTorch)4m
   • What happens inside the pipeline function? (TensorFlow)5m
   • Instantiate a Transformers model (PyTorch)3m
   • Instantiate a Transformers model (TensorFlow)3m
   • Tokenizers Overview0m
   • Word-based tokenizers2m
   • Character-based tokenizers3m
   • Subword-based tokenizers3m
   • The tokenization pipeline3m
   • Batching inputs together (PyTorch)2m
   • Batching inputs together (TensorFlow)2m
   • Hugging Face Datasets overview (Pytorch)3m
   • Hugging Face Datasets overview (Tensorflow)3m
   • Preprocessing sentence pairs (PyTorch)3m
   • Preprocessing sentence pairs (TensorFlow)3m
   • What is dynamic padding?3m
   • The Trainer API3m
   • Keras introduction2m
   • Fine-tuning with TensorFlow5m
   • Learning rate scheduling with TensorFlow4m
   • TensorFlow Predictions and metrics4m
   • Write your training loop in PyTorch4m
   • Supercharge your PyTorch training loop with Accelerate3m
   • Navigating the Model Hub3m
   • Managing a repo on the Model Hub7m
   • The Push to Hub API (PyTorch)5m
   • The Push to Hub API (TensorFlow)8m
   • Loading a custom dataset3m
   • Slice and dice a dataset 🔪3m
   • Datasets + DataFrames = ❤️2m
   • Saving and reloading a dataset3m
   • Memory mapping & streaming3m
   • Uploading a dataset to the Hub2m
   • Text embeddings & semantic search3m
   • Training a new tokenizer6m
   • Why are fast tokenizers called fast?1m
   • Fast tokenizer superpowers3m
   • Inside the Token classification pipeline (PyTorch)3m
   • Inside the Token classification pipeline (TensorFlow)3m
   • Inside the Question answering pipeline (PyTorch)3m
   • Inside the Question answering pipeline (TensorFlow)3m
   • What is normalization?5m
   • What is pre-tokenization?2m
   • Byte Pair Encoding Tokenization5m
   • WordPiece Tokenization3m
   • Unigram Tokenization8m
   • Building a new tokenizer5m
   • Data processing for Token Classification3m
   • Data processing for Masked Language Modeling2m
   • What is perplexity?2m
   • What is domain adaptation?1m
   • Data processing for Translation2m
   • What is the BLEU metric?4m
   • Data processing for Summarization2m
   • What is the ROUGE metric?4m
   • Data processing for Causal Language Modeling4m
   • Using a custom loss function3m
   • Data processing for Question Answering2m
   • The Post processing step in Question Answering (PyTorch)3m
   • The Post processing step in Question Answering (TensorFlow)3m
   • Data Collators: A Tour6m
   • What to do when you get an error?2m
   • Using a debugger in a notebook3m
   • Using a debugger in a terminal4m
   • Asking for help on the forums3m
   • Debugging the Training Pipeline (PyTorch)4m
   • Debugging the Training Pipeline (TensorFlow)7m
   • Writing a good issue3m
   • Hugging Face Course Event Day 14h 43m
   • Hugging Face Course Event Day 21h 59m
   • Hugging Face Course Event Day 2 - Part 21h 52m