llama3 implementation one matrix multiplication at a time

Minimal, clean code for the Byte Pair Encoding (BPE) algorithm commonly used in LLM tokenization.

Inference Llama 2 in one file of pure C

LLM training in simple, raw C/CUDA