🎮 Game

An in-game TinyStories LLM chat that lets the player enter a prompt and get locally generated text in a UI panel. Under the hood it loads a packaged model and tokenizer from mounted resources, runs a GPT‑Neo style forward pass and a greedy generator on a worker thread, and validates shapes and outputs against bundled reference data. The code includes tensor utilities, a GPT‑2 byte BPE tokenizer, binary model and reference loaders, and parity test harnesses with detailed diagnostics.