Completed
- 1:53:11 Multi-Head Attention
Class Central Classrooms beta
YouTube videos curated by Class Central.
Classroom Contents
Code Your Own Llama 4 LLM from Scratch - Full Course
Automatically move to the next video in the Classroom when playback concludes
- 1 - 0:00:00 Introduction to the course
- 2 - 0:00:15 Llama 4 Overview and Ranking
- 3 - 0:00:26 Course Prerequisites
- 4 - 0:00:43 Course Approach for Beginners
- 5 - 0:01:27 Why Code Llama from Scratch?
- 6 - 0:02:20 Understanding LLMs and Text Generation
- 7 - 0:03:11 How LLMs Predict the Next Word
- 8 - 0:04:13 Probability Distribution of Next Words
- 9 - 0:05:11 The Role of Data in Prediction
- 10 - 0:05:51 Probability Distribution and Word Prediction
- 11 - 0:08:01 Sampling Techniques
- 12 - 0:08:22 Greedy Sampling
- 13 - 0:09:09 Random Sampling
- 14 - 0:09:52 Top K Sampling
- 15 - 0:11:02 Temperature Sampling for Controlling Randomness
- 16 - 0:12:56 What are Tokens?
- 17 - 0:13:52 Tokenization Example: "Hello world"
- 18 - 0:14:30 How LLMs Learn Semantic Meaning
- 19 - 0:15:23 Token Relationships and Context
- 20 - 0:17:17 The Concept of Embeddings
- 21 - 0:21:37 Tokenization Challenges
- 22 - 0:22:15 Large Vocabulary Size
- 23 - 0:23:28 Handling Misspellings and New Words
- 24 - 0:28:42 Introducing Subword Tokens
- 25 - 0:30:16 Byte Pair Encoding BPE Overview
- 26 - 0:34:11 Understanding Vector Embeddings
- 27 - 0:36:59 Visualizing Embeddings
- 28 - 0:40:50 The Embedding Layer
- 29 - 0:45:31 Token Indexing and Swapping Embeddings
- 30 - 0:48:10 Coding Your Own Tokenizer
- 31 - 0:49:41 Implementing Byte Pair Encoding
- 32 - 0:52:13 Initializing Vocabulary and Pre-tokenization
- 33 - 0:55:12 Splitting Text into Words
- 34 - 1:01:57 Calculating Pair Frequencies
- 35 - 1:06:35 Merging Frequent Pairs
- 36 - 1:10:04 Updating Vocabulary and Tokenization Rules
- 37 - 1:13:30 Implementing the Merges
- 38 - 1:19:52 Encoding Text with the Tokenizer
- 39 - 1:26:07 Decoding Tokens Back to Text
- 40 - 1:33:05 Self-Attention Mechanism
- 41 - 1:37:07 Query, Key, and Value Vectors
- 42 - 1:40:13 Calculating Attention Scores
- 43 - 1:41:50 Applying Softmax
- 44 - 1:43:09 Weighted Sum of Values
- 45 - 1:45:18 Self-Attention Matrix Operations
- 46 - 1:53:11 Multi-Head Attention
- 47 - 1:57:55 Implementing Self-Attention
- 48 - 2:10:40 Masked Self-Attention
- 49 - 2:37:09 Rotary Positional Embeddings RoPE
- 50 - 2:38:08 Understanding Positional Information
- 51 - 2:40:58 How RoPE Works
- 52 - 2:49:03 Implementing RoPE
- 53 - 2:56:47 Feed-Forward Networks FFN
- 54 - 2:58:50 Linear Layers and Activations
- 55 - 3:02:19 Implementing FFN
