LLM Lecture Series: How Large Language Models Work Round 3

Share This Post

7 Sessions (2 hours each)

Understanding the architecture, training, and computational mechanics of large language models
and Transformers.

ICT professionals, engineers, developers, and technically curious professionals and graduates.

Module 1 – Introduction to LLMs (2 lectures/4 hours)

AI and ML foundations, language models, the rise of Transformers, emergent properties, and the AGI debate.

Module 2 – Preliminary Concepts (1 lecture/2 hours)

Neural networks, the error back propagation algorithm, sequence modelling, loss functions, softmax, NER, and key NLP tasks.

Module 3 – Work Embeddings (1 lecture/2 hours)

Word2Vec, GloVe, FastText, contextual embeddings, and practical embedding workflows.

Module 4 – How ChatGPT Works and is Trained (2 lectures/4 hours)

GPT-style Transformers in detail: positional encodings, Q-K-V attention, residuals, normalisation, projection layers, model training, fine-tuning, hallucination, and the way forward.

Full slide deck for all lectures (made available to registered attendees)
Slides include worked diagrams and step-by-step breakdowns of the Transformer and Q-K-V computations
Curated references for further study

Dates