Learn about transformers, the go-to architecture for NLP and computer vision tasks.
Overview
Syllabus
Introduction
- Natural language processing with transformers
- How to use the exercise files
- How transformers are used in NLP
- Transformers in production
- Transformers history
- Challenge: BERT model sizes
- Solution: BERT model sizes
- Bias in BERT
- How was BERT trained?
- Transfer learning
- Transformer: Architecture overview
- BERT model and tokenization
- Positional encodings and segment embeddings
- Tokenizers
- Self-attention
- Multi-head attention and feedforward network
- BERT and text classification
- The Datasets library
- Overview of IMDb dataset
- Using a tokenizer
- Tiny IMDb
- A training run
- Additional training runs
Taught by
Jonathan Fernandes