Stat 359 course materials
Lecture notes can be found on the course Canvas website.
| Lecture | Date | Material | Readings |
|---|---|---|---|
| Week 1, Thursday | January 4 | Introduction | Perplexity, Perplexity 2, Linear Models |
| Week 2, Tuesday | January 9 | Attention | Attention is All You Need, Attention Mechanisms, Attention with code, The Illustrated Transformer |
| Week 2. Thursday | January 11 | Transformers | Intro to Transformers, Discussion, Blog post, Skip connections, Layer normalization, Byte-Pair Encoding |
| Week 3, Tuesday | January 16 | Coding a transformer | Code example |
| Week 3, Thursday | January 18 | BERT, GPT, LLAMA | Annotated GPT 2, GPT-2 paper, Adversarial attacks on GPT-2 LLAMA Paper, BERT, Understanding BERT |
| Week 4, Tuesday | January 23 | Prompt Tuning, chain of thought, hindsight chain of thought, backwards chain of thought, Graph of Thought, Tree of Thought, Training Chain-of-Thought via Latent-Variable Inferenc, prompt engineering | Language Models are Few Shot Learners, Zero shot chain of thought, LLMs are human-level prompt engineers, Tree of Thought, Chain of verification, Promptbreeder, Prompt Engineering Guide, Open-AI Prompting Guide |
| Week 4, Thursday | January 25 | Fine tuning, tool use, parameter-efficient fine tuning, LORA, Instruction Tuning (SFT), Neftune, quantization | LORA, PEFT, Quantization, Quantization Blog, Alpaca, ToolEMU, NEFTune, Tool Use code, Model Calibration |
| Week 5, Tuesday | January 30 | Hugging face, fine tuning LLAMA | Mistral Fine Tuning, LLAMA fine tuning |
| Week 5, Thursday | February 1 | No class | |
| Week 6, Tuesday | February 6 | RAG, when to use RAG vs SFT, lexacagraphical vs semantic search, sentence transformers, Retrieval transformers and long term memory in transformers, RAG Code. | RAG, RAG code, Sentence Transformers, Retrieval Transformers, Improving Neural Language Models with a Continuous Cache |
| Week 6, Thursday | February 8 | ChatGPT and RLHF, rejection sampling, DPO, Gopher Cite | Chain of hindsight |
| Week 7, Tuesday | February 13 | Asking questions about images. Conditional layer norm, FILM, CLIP, BLIP, LAVA | |
| Week 7, Thursday | February 15 | Diffusion models, DDPM, classifier-free guidance | |
| Week 8, Tuesday | February 20 | Stable Diffusion, tuning stable diffusion, Brian’s code | |
| Week 8, Thursday | February 22 | Frontiers, using LLMs to help diffusion models by planning out images. Instance recognition and inserting new objects %s tricks. Consistency models, SD Edit, Diffusion in robotics. | |
| Week 9, Tuesday | February 27 | Presentations | |
| Week 9, Thursday | February 29 | Presentations | |
| Week 10, Tuesday | March 5th | Presentations |
| Project title | Date released | Due date |
|---|---|---|
| Assignment 1 | Jan 9 | Jan 25 |
| Assignment 2 | Jan 30 | Feb 15 |
| Final presentation topic proposals | Feb 15 | |
| Final presentations | Feb 27 |