Working with Encoder, Decoder, and LLMs
This Jupyter Notebook provides a hands-on guide to text generation, progressing from foundational concepts to modern Large Language Models (LLMs). It begins by demonstrating how to build a classic sequence-to-sequence (Seq2Seq) translation model from scratch using LSTMs, comparing versions with and without an attention mechanism. The tutorial then advances to using the pre-trained T5 model for tasks like summarization and translation through simple input prefixes. Finally, it explores various prompt engineering techniques for powerful LLMs, showcasing prefix prompts with Facebook's OPT, instruction-following with the Qwen3 model, and offering a comparison to ChatGPT, effectively covering the evolution from building models to prompting them.
