GuppyLM

GuppyLM is a small language model with 8.7M parameters designed to simulate a fish character named Guppy. It is trained from scratch in about 5 minutes on a single GPU using synthetic conversational data covering 60 different aquarium-related topics. The model runs entirely locally in a browser via WebAssembly or can be interacted with via Python scripts and Colab notebooks. It demonstrates how to build and train a simple transformer LLM to understand foundational transformer architecture and training pipelines without requiring large-scale resources.