Tiny LLM built from scratch in PyTorch: loads GPT-2, Llama/SmolLM2, Qwen2.5 and Pythia weights, LoRA fine-tuning, KV cache, and a Streamlit chatbot.