Complete LLM chatbot implementation, offering tokenization, pretraining, finetuning, evaluation, inference, and a simple web UI. Runs on a single 8xH100 node, features a hackable codebase, dependency-lite design, and automates the full workflow via scripts.






































































