Events / Boston

Building a Transformer LM

A hands-on reading/coding group working through Stanford CS336 Assignment 1, building a transformer's tokenizer, attention layers, and full architecture from scratch.

Thu, Jul 30, 10:00 PMBoston Public Library · Boston · MA

Why it matters

Understanding how tokenization and attention actually work under the hood demystifies why certain prompts and contexts cost what they do.

The tokenmaxxing angle

The session literally implements a tokenizer from scratch, the exact mechanism that turns text into billable tokens, plus attention, which drives context-window compute cost.

From the organizers

Group is working through Stanford CS336 (Language Modeling from Scratch) Assignment 1, building tokenization, attention, and full transformer architecture, hosted by Mark R. at Boston Public Library.