Module 2 begins: just enough maths. In Lecture 11 of the LLM Masterclass: why everything inside an LLM is numbers, vectors as lists of numbers and arrows, the dot product (chai · tea = 1.47, chai · car = 0.345), matrices and their shape, why a matrix multiplication is really many dot products, and all of it in NumPy with the @ operator. 💻 Code for this lecture (branch lecture-11): https://github.com/programmingline/chotugpt/tree/lecture-11 Lecture 11 of 32 — LLM Masterclass · Module 2: Just enough maths ⚠️ ChotuGPT is a simplified teaching model. The word vectors (chai, tea, car) and the small 'hot drink? / vehicle?' matrix are toy numbers for teaching only. In this lecture: • Everything becomes numbers: token IDs (lecture 2) and word vectors (lecture 5) • A vector is a list of numbers; its length is its number of dimensions • The dot product: multiply matching numbers and add — big when two vectors point the same way • Why cosine similarity divides by the lengths • Matrices: a grid of numbers with a shape (rows × columns); learned weights live in matrices • Matrix × vector: each output is one dot product; matrix × matrix: row times column, inner sizes must match • NumPy: @ for matrix multiplication, words @ words.T gives every pair's dot product at once • Why GPUs matter, and how attention scores are dot products (lecture 18) Chapters: 00:00 Lecture 10 quiz answer 00:12 Module 2: just enough maths 00:24 Everything becomes numbers 00:40 Vectors 01:05 Starting to click? 01:10 The dot product 01:43 Matrices and shape 02:02 Matrix multiplication = many dot products 02:42 NumPy: the @ operator 03:07 Quiz 03:27 Why this matters: Transformers and GPUs 03:47 Recap & next lecture ✅ Lecture 10 quiz answer: B, training on human feedback to follow instructions. The big change was not size. ❓ Quiz: A matrix multiplication is really many… A) Additions B) Dot products C) Divisions. Comment your answer. I reveal it in Lecture 12! 👍 Like, subscribe and comment your quiz answer. ⏮ Previous: Lecture 10 — From GPT-3 to ChatGPT: the modern era ⏭ Next: Lecture 12 — A neural network from scratch Sources: NumPy documentation: numpy.matmul and the @ operator (numpy.org/doc); G. Strang, "Introduction to Linear Algebra" (Wellesley-Cambridge Press); toy word vectors from Lecture 5 of this course. #LLM #LinearAlgebra #NumPy
The information provided is not trading advice. kdj.com does not assume any responsibility for any investments made based on the information provided in this article. Cryptocurrencies are highly volatile and it is highly recommended that you invest with caution after thorough research!
If you believe that the content used on this website infringes your copyright, please contact us immediately (info@kdj.com) and we will delete it promptly.