Reverse Engineering Large Language Models | Build Your Own LLM Workshop #2 [Refreshed]
Reverse Engineering LLMs: print(), summary(), and a Roadmap to Building GPT-2-Style Transformers. Part of a Build your own LLM workshop. =========== LINKS Justin's twitter: https://x.com/JustinAngel Workshop overview: https://go.justinangel.ai/substack Deck: https://go.justinangel.ai/deck Google Drive: https://go.justinangel.ai/drive Code exercise: https://go.justinangel.ai/code-2 =========== CHAPTERS 00:00 Workshop Intro 01:14 Load GPT2 in Colab 01:28 Model Size and Params 02:14 Print and Summary 03:49 Module Tree Breakdown 04:39 Roadmap from Components 05:34 Function Call Graph 06:43 Hugging Face Exercise 08:48 Ablation Tests Explained 10:10 LLMs Grown Not Built 11:28 Full Roadmap and Deliverables 13:13 Interactive Demos and Models 14:52 Next Up Perceptrons =============== ABOUT THIS TALK In this episode of the Build Your Own Large Language Model workshop, the presenter begins Section 2, “Reverse Engineering LLMs,” using GPT-2 in a Colab notebook to extract an inference-time roadmap by calling print(model) and torchinfo summary. They note GPT2-small has about 124M parameters and show how the architecture reveals repeated GPT2 blocks (12 layers) and components like embeddings, dropout, LayerNorm, Conv1D/linear layers, attention, MLPs, and Gelu, plus function-level elements like initialization and softmax. They demonstrate repeating the process on a different Hugging Face model (TinyLlama/LlamaModel), highlighting shared parts and differences like RMSNorm and rotary embeddings, and encourage viewers to try other models. The episode introduces ablation tests as “taking away” components to see what works, shares quotes about LLMs being “grown,” and previews upcoming sections starting with perceptrons, activations, MLPs, and training fundamentals like loss and backprop.

Perceptrons: wx+b | Build Your Own LLM Workshop #3

Using Large Language Models | Build Your Own LLM Workshop #1
![Activation Functions: ReLU, GELU | Build Your Own LLM Workshop #4 [Refreshed]](https://i.ytimg.com/vi/ozvUxyDJvl4/hqdefault.jpg?sqp=-oaymwEjCNACELwBSFryq4qpAxUIARUAAAAAGAElAADIQj0AgKJDeAE=&rs=AOn4CLAGxA1zsnQyWMmbnO_J53k_-zwS8A)
Activation Functions: ReLU, GELU | Build Your Own LLM Workshop #4 [Refreshed]

What did Anthropic do?! (Opus 5)

🎙️Streaming JSON SAX Parsing for Yosys Netlists
![Perceptrons: wx+b | Build Your Own LLM Workshop #3 [Refreshed]](https://i.ytimg.com/vi/4afEe4Mk-3c/hqdefault.jpg?sqp=-oaymwEjCNACELwBSFryq4qpAxUIARUAAAAAGAElAADIQj0AgKJDeAE=&rs=AOn4CLABRA4a8JYo3MlxPO7zKlOkLju_DQ)
Perceptrons: wx+b | Build Your Own LLM Workshop #3 [Refreshed]

Starship Flight 13 - Where's The Kaboom?

What We Didn't Cover | Build Your Own LLM Workshop #23

Alexander Mercouris: NATO gerät bald in Panik und riskiert Krieg mit Russland

China Could Replace Traditional Solar Panels Forever With a New Material

Debt Crisis 2.0: The Euro is Facing its Next Collapse – Before the Storm with Tichy and Kolbe

Activation Functions: ReLU, GELU | Build Your Own LLM Workshop #4

The End of Consoles

STUDY: GERMANY IS WAKING UP | Climate skeptics, green loss of prosperity & CO2 tax

翻譯 Andrej Karpathy Deep Dive Into LLMs Like ChatGPT

I Tested Meta's New AI Glasses With the Man Who Built Them

Tesla Semi Shocks Skeptics — Survives 82,000lbs in Brutal -40°C

Programming the Apollo Guidance Computer | Blockly Summit 2026

