base-llm

Why we included this project

This is a code-first tutorial that walks developers from classical NLP foundations up to modern large language model practice, organized as a progression rather than scattered notes. It covers tokenization and word vectors, RNN/LSTM, the Transformer and attention, then BERT/GPT/T5, before moving into hands-on chapters that build a Llama2-style model from scratch, apply LoRA and QLoRA fine-tuning, and cover RLHF. Working engineers will get the most from the deployment section, which shows how to serve a fine-tuned model with FastAPI, containerize it with Docker Compose, and set up CI/CD, so the material carries through to real production work. Practical projects like text classification and NER round it out, along with chapters on model safety and multimodal models, giving learners a broad but coherent view of the field.