Lee Miller

WordLlama. Things you can do with the token embeddings of an LLM

string-noise. Efficiently simulate different noise types for training-time augmentations

PennyLM. Training scripts for Penny-1.7B

nopilot.nvim. Neovim plugin for interaction with LLMs

auto-prompt-opt-dspy-cross-encoders. Python

WordLlamaDetect. Tokenization based language detection

splade-easy. A simple retriever for SPLADE sparse encodings

GrammarLee. Lee's grammar utility

SelectStuff. Browser extension select -> LLM indexing

CnakeCharmer. A "living dataset" of python -> cython pair implementations