Building Large Language Models (LLMs) from Scratch

Building Large Language Models (LLMs) from Scratch – Full Series

عدد الدروس : 43 عدد ساعات الدورة : 30:57:00 شهادة معتمدة : نعم التسجيل في الدورة للحصول على شهادة

للحصول على شهادة

1- التسجيل
2- مشاهدة الكورس كاملا
3- متابعة نسبة اكتمال الكورس تدريجيا
4- بعد الانتهاء تظهر الشهادة في الملف الشخصي الخاص بك

Master the fundamentals and advanced techniques of building LLMs from scratch, including tokenization, embeddings, transformers, attention mechanisms, and full architecture coding.

قائمة الدروس

1 - Lecture 1: Building LLMs from scratch: Series introduction

2 - Lecture 2: Large Language Models (LLM) Basics

3 - Lecture 3: Pretraining LLMs vs Finetuning LLMs

4 - Lecture 4: What are transformers?

5 - Lecture 5: How does GPT-3 really work?

6 - Lecture 6: Stages of building an LLM from Scratch

7 - Lecture 7: Code an LLM Tokenizer from Scratch in Python

8 - Lecture 8: The GPT Tokenizer: Byte Pair Encoding

9 - Lecture 9: Creating Input-Target data pairs using Python DataLoader

10 - Lecture 10: What are token embeddings?

11 - Lecture 11: The importance of Positional Embeddings

12 - Lecture 12: The entire Data Preprocessing Pipeline of Large Language Models (LLMs)

13 - Lecture 13: Introduction to the Attention Mechanism in Large Language Models (LLMs)

14 - Lecture 14: Simplified Attention Mechanism - Coded from scratch in Python | No trainable weights

15 - Lecture 15: Coding the self attention mechanism with key, query and value matrices

16 - Lecture 16: Causal Self Attention Mechanism | Coded from scratch in Python

17 - Lecture 17: Multi Head Attention Part 1 - Basics and Python code

18 - Lecture 18: Multi Head Attention Part 2 - Entire mathematics explained

19 - Lecture 19: Birds Eye View of the LLM Architecture

20 - Lecture 20: Layer Normalization in the LLM Architecture

21 - GELU Activation Function in the LLM Architecture

22 - Shortcut connections in the LLM Architecture

23 - Coding the entire LLM Transformer Block

24 - Coding the 124 million parameter GPT-2 model

25 - Coding GPT-2 to predict the next token

26 - Measuring the LLM loss function

27 - Evaluating LLM performance on real dataset | Hands on project | Book data

28 - Coding the entire LLM Pre-training Loop

29 - Temperature Scaling in Large Language Models (LLMs)

30 - Top-k sampling in Large Language Models

31 - Saving and loading LLM model weights using PyTorch

32 - Loading pre-trained weights from OpenAI GPT-2

33 - Introduction to LLM Finetuning | Python Coding with hands-on-example

34 - Dataloaders in LLM Classification Finetuning | Python Coding | Hands on LLM project

35 - Coding the model architecture for LLM classification fine-tuning

36 - Coding a fine-tuned LLM spam classification model | From Scratch

37 - Introduction to LLM Instruction Fine-tuning | Loading Dataset | Alpaca Prompt format

38 - Data Batching in LLM instruction fine-tuning | Hands on project | Live Python coding

39 - Dataloaders in Instruction Fine-tuning

40 - Instruction fine-tuning: Loading pre-trained LLM weights

41 - LLM fine-tuning training loop | Coded from scratch

42 - Evaluating fine-tuned LLM using Ollama

43 - Build LLMs from scratch 20 minutes summary

عن الدورة

This course, Building Large Language Models (LLMs) from Scratch – Full Series, provides a comprehensive, step-by-step guide to understanding and creating LLMs. Starting with the basics of LLMs, the series covers pretraining vs. finetuning, transformers, and the inner workings of GPT-3.

Students will learn the stages of building an LLM, from coding tokenizers in Python to creating input-target data pairs, understanding token and positional embeddings, and implementing the entire data preprocessing pipeline. Advanced topics include attention mechanisms, self-attention, causal attention, multi-head attention, and the mathematics behind them.

The course also explains LLM architectural components such as layer normalization, GELU activation, shortcut connections, and the complete transformer block. Practical coding exercises throughout the series enable learners to implement all components from scratch, reinforcing both conceptual understanding and hands-on skills.

By the end of the series, learners will have a deep understanding of LLM construction, transformer mechanics, and the ability to build, test, and optimize large language models for real-world applications. This course is ideal for AI researchers, ML engineers, and developers looking to master state-of-the-art NLP model creation.