Normcore LLM Reads

normcore-llm.md

Anti-hype LLM reading list

Goals: Add links that are reasonable and good explanations of how stuff works. No hype and no vendor content if possible. Practical first-hand accounts of models in prod eagerly sought.

Foundational Concepts

Pre-Transformer Models

• The Illustrated Word2vec - A Gentle Intro to Word Embeddings in Machine Learning (YouTube)
• Transformers as Support Vector Machines
• Survey of LLMS
• Deep Learning Systems
• Fundamental ML Reading List

Building Blocks

• What are embeddings
• Concepts from Operating Systems that Found their way into LLMS
• Talking about Large Language Models
• Language Modeling is Compression
• Vector Search - Long-Term Memory in AI
• Eight things to know about large language models
• The Bitter Lesson
• The Hardware Lottery
• The Scaling Hypothesis
• Tokenization
• LLM Course

Foundational Deep Learning Papers (in semi-chronological order)

• Seq2Seq
• Attention is all you Need
• BERT
• GPT-1
• Scaling Laws for Neural Language Models
• T5
• GPT-2: Language Models are Unsupervised Multi-Task Learners
• InstructGPT: Training Language Models to Follow Instructions
• GPT-3: Language Models are Few-Shot Learners

The Transformer Architecture

• Transformers from Scratch
• Transformer Math
• Five Years of GPT Progress
• Lost in the Middle: How Language Models Use Long Contexts

Attention

• Self-attention and transformer networks
• Attention
• Understanding and Coding the Attention Mechanism
• Attention Mechanisms
• Keys, Queries, and Values

GPT

• What is ChatGPT doing and why does it work
• My own notes from a few months back.
• Karpathy's The State of GPT (YouTube)
• OpenAI Cookbook

Significant OSS Models

• Llama2
• Mistral7B
  • Mixtral
• Phi2
• Falcon7B

LLMs in 2023

• Catching up on the weird world of LLMS
• How open are open architectures?
• Building an LLM from Scratch
• Large Language Models in 2023 and Slides
• Timeline of Transformer Models
• Large Language Model Evolutionary Tree

Training Data

• What's in my Big Data
• "The “it” in AI models is the dataset."
• Extracting Training Data from ChatGPT

Pre-Training

• Why host your own LLM?
• How to train your own LLMs
• Hugging Face Resources on Training Your Own
• Training Compute-Optimal Large Language Models
• Opt-175B Logbook

RLHF and DPO

• RLHF
  • Supervised Fine-tuning
  • How Abilities in LLMs Are Affected by SFT
• Instruction-tuning for LLMs: Survey
• Direct Preference Optimization: Your Language Model is Secretly a Reward Model
• RLHF and DPO Compared

Fine-Tuning and Compression

• The Complete Guide to LLM Fine-tuning
• LLaMAntino: LLaMA 2 Models for Effective Text Generation in Italian Language - Really great overview of SOTA fine-tuning techniques
• On the Structural Pruning of Large Language Models
• Quantiztion
  • A Gentle Introduction to 8-bit matrix multiplication
  • Which Quantization Method is Right for You?
  • Survey of Quantization for Inference
• PEFT
  • Fine-tuning with LoRA and QLoRA
  • Adapters
  • Motivation for Parameter-Efficient Fine-tuning

Small and Local LLMs

• How is LlamaCPP Possible?
• How to beat GPT-4 with a 13-B Model
• Efficient LLM Inference on CPUs
• Tiny Language Models Come of Age
• Efficiency LLM Spectrum
• TinyML at MIT

Deployment and Production

• Building LLM Applications for Production
• Challenges and Applications of Large Language Models
• All the Hard Stuff Nobody talks about when building products with LLMs
• Scaling Kubernetes to run ChatGPT
• Numbers every LLM Developer should know
• Against LLM Maximalism
• A Guide to Inference and Performance
• (InThe)WildChat: 570K ChatGPT Interaction Logs In The Wild
• The State of Production LLMs in 2023
• Machine Learning Engineering for successful training of large language models and multi-modal models.
• Fine-tuning RedPajama on Slack Data

LLM Inference and K-V Cache

• LLM Inference Performance Engineering: Best Practices
• How to Make LLMs go Fast
• Transformer Inference Arithmetic
• Which serving technology to use for LLMs?
• Speeding up the K-V cache
• Large Transformer Model Inference Optimization

Prompt Engineering and RAG

• On Prompt Engineering
• Prompt Engineering Versus Blind Prompting
• Building RAG-Based Applications for Production
• Full Fine-Tuning, PEFT, or RAG?
• Prompt Engineering Guide

GPUs

• The Best GPUS for Deep Learning 2023
• Making Deep Learning Go Brr from First Principles
• Everything about Distributed Training and Efficient Finetuning
• Training LLMs at Scale with AMD MI250 GPUs
• GPU Programming

Evaluation

• Evaluating ChatGPT
• ChatGPT: Jack of All Trades, Master of None
• What's Going on with the Open LLM Leaderboard
• Challenges in Evaluating AI Systems
• LLM Evaluation Papers
• Evaluating LLMs is a MineField

Eval Frameworks

• HELM
• LM Eval Harness
• LmSys Chatbot Arena

UX

• Generative Interfaces Beyond Chat (YouTube)
• Why Chatbots are not the Future
• The Future of Search is Boutique
• As a Large Language Model, I
• Natural Language is an Unnatural Interface

What's Next?

Thanks to everyone who added suggestions on Twitter, Mastodon, and Bluesky.

Transformer Math 101

We present basic math related to computation and memory usage for transformers

Why Chatbots Are Not the Future

Posts from eugene yan are also pretty good read.

Hi, thank you for your great work!

Training Your Own

I wonder if we can add huggingface/llm_training_handbook: An open collection of methodologies to help with successful training of large language models. to this section?

It seems that the gzip approach, altough really cool, was 'optimistic' and thus overhyped, see: https://kenschutte.com/gzip-knn-paper/ (basiccaly they confused k in k-nn and top-k accuracy, reporting top-2 accuracy). More recent studies found that it is, as expected, on 'bag of words' performance level Gzip versus bag-of-words for text classification.

I don't know if you intend to (or are even interested) but I am on the look out for "usecases for normies".

Do you think it's worth annotating with dates of the articles / papers / videos?

Why Chatbots Are Not the Future

Great one!

It seems that the gzip approach, altough really cool, was 'optimistic' and thus overhyped, see: https://kenschutte.com/gzip-knn-paper/ (basiccaly they confused k in k-nn and top-k accuracy, reporting top-2 accuracy). More recent studies found that it is, as expected, on 'bag of words' performance level Gzip versus bag-of-words for text classification.

I don't know if you intend to (or are even interested) but I am on the look out for "usecases for normies".

Yeah that was my read on it as well but I'm also very interested in it as a general theoretical approach and baseline, even if this particular implementation doesn't work.

Do you think it's worth annotating with dates of the articles / papers / videos?

Maybe would be helpful but I explicitly picked stuff that I thought wouldn't age and/or where the recency didn't matter because the fundamentals are timeless.

Patterns for Building LLM-based Systems & Products

In my opinion, this is a super in depth article that covers many of the categories and deserves a place in the reading list.

Against LLM maximalism (disclaimer: I work at Explosion)

Patterns for Building LLM-based Systems & Products In my opinion, this is a super in depth article that covers many of the categories and deserves a place in the reading list.

detailed blog

The Illustrated Transformer - maybe redundant from some of the other transformers content here, but is very well-written and strikes a good balance between prose and visual aids.

ChatGPT Sucks at Being a Testing Expert
https://www.satisfice.com/download/appendix-chatgpt-sucks-at-being-a-testing-expert?wpdmdl=487569

This is a careful analysis of an attempt to demonstrate ChatGPT’s usefulness to help testers.

I liked this article a lot: https://lilianweng.github.io/posts/2023-06-23-agent/

My writing as an extension to "Why you should host your LLM?" article, with some adding on operation perspective: https://tulip4attoo.substack.com/p/why-you-should-host-your-llm-from

Maybe Foundational Papers should include the first instruction-tuned model FLAN (no RLHF)：
Finetuned Language Models Are Zero-Shot Learners: https://arxiv.org/abs/2109.01652

I gave a talk about prompt engineering for normal people and turned it into a pretty decent article, might be useful for the list too? https://timbornholdt.com/blog/prompt-engineering-how-to-think-like-an-ai

The Stochastic Parrots paper presents many things that anyone should be cognisant of when deciding whether or not to use an LLM:

Bender, Emily M., Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. 2021. On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜. In Proceedings of FAccT 2021, pp.610-623.

We are exploring the landing and commercialization scenarios of AI Agent at https://askgen.ie, and currently we think customer support is a good scenario

@livc Are you guys hiring?

LLMs for designing 2D game levels: Level Generation Through Large Language Models
LLMs for Neural Architecture Search : LLMatic: Neural Architecture Search via Large Language Models and Quality-Diversity Optimization

I found "Five years of GPT progress" to be a useful overview of the influential papers on GPT.
https://finbarr.ca/five-years-of-gpt-progress/
May work as a high-level summary for "Foundational Papers" section.

p.s. Thank you for compiling the list!

Hi! This is great! Would be even more useful to write the year/month of publication next to each item, to get a sense of which links are more up-to-date and which are more historical.

I really like this list, sad I just discovered this 😎 .

I am not sure if this would complement your Background section, but I wrote this as a primer on LLMs last month: https://willthompson.name/what-we-know-about-llms-primer.

But I don't know, might not be very orthogonal to your other sources here 🤷 .

An overview of vector database. The author highlight the differences between the various vector databases out there as visually as possible.

https://thedataquarry.com/posts/vector-db-1/

An overview of vector database. The author highlight the differences between the various vector databases out there as visually as possible.

https://thedataquarry.com/posts/vector-db-1/

This is really helpful, thank you!

GPT in 60 Lines of NumPy https://jaykmody.com/blog/gpt-from-scratch/

I keep coming back to this list. However I feel like it miss a good discussion about current stuff not working. I keep failling to implement working stuff, despite lenghty theoretical works, and when I scratch the veneer I keep getting the same answer: "technology is not ready yet".

Also: Bash One-Liners for LLMs