Ecole Centrale Paris, MIT. Twin of @shervinea. New book on Transformers & Large Language Models: https://teapawt.net/t.co/5b89hbceJf
Bay Area
Joined August 2014
- Tweets82
- Following77
- Followers5.1K
- Likes531
Our Stanford class "CME 295: Transformers & Large Language Models" is back for a new edition and Lecture 1 is now on YouTube!
Announcing the VIP Cheatsheet for Stanford's CME 295 Transformers & Large Language Models class.
Topics include:
- Transformers: self-attention, architecture, variants, optimization techniques (sparse attention, low-rank attention, flash attention)
- LLMs: prompting, finetuning (SFT, LoRA), preference tuning, optimization techniques (mixture of experts, distillation, quantization)
- Applications: LLM-as-a-judge, RAG, agents, reasoning models (train-time and test-time scaling from DeepSeek-R1)
Link to PDF: github.com/afshinea/stanford…
9
1,025
This Spring, my twin brother @shervinea and I will be teaching a new class at Stanford called "Transformers & Large Language Models" (CME 295).
The goal of this class is to understand where LLMs come from, how they are trained, and where they are most used. We will also explore more recent topics such as reasoning models and agentic workflows.
We will be posting updates on the class website and uploading some relevant material in the coming weeks.
For reference, a large portion of the class is based on our recent book "Super Study Guide: Transformers & Large Language Models".
Stay tuned & happy learning!
「Super Study Guide: Transformer と大規模言語モデル」日本語版が発売されました。250 ページにわたり約 600 点のカラーの図を交えて Transformer と LLM の詳細を解説しています。
Announcing "Super Study Guide: Transformers & Large Language Models", a 250-page book with ~600 colored illustrations covering the concepts of the Stanford workshop that Shervine and I are teaching at this summer.
Topics include:
- Transformers (how they work, detailed examples)
- LLMs (pretraining, prompt engineering, finetuning, preference tuning, optimizations such as quantization)
- Applications (RAG, translation, etc.)
Book: amazon.com/dp/B0DC4NYLTN/
Thank you @EngineerHaydn for your detailed review of our new book!
youtube.com/watch?v=e6lmXQh0…
Announcing Super Study Guide: Algorithms and Data Structures.
Ideal for visual learners who want to brush up on core CS skills. Topics: arrays/strings, queues/stacks, hash tables, graphs, trees, sorting and search.
Web: superstudy.guide
Book: amazon.com/dp/B0B2TW6896
Earlier this summer, my twin brother and I decided to do a road trip between our two alma maters: MIT and Stanford. We recorded everything in case you would like to see what this (very) long trip looked like.
Full video on YouTube: youtu.be/4Nl61iyCySc
cc: @MIT, @Stanford
Ideal to prepare for interviews: illustrated Data Science study guides covering MIT’s 15.003 Software Tools class.
Topics: data retrieval with SQL, data manipulation with R and Python, tips with Bash and Git.
Web: mit.edu/~amidi/teaching/data…
PDF version: github.com/shervinea/mit-15-…
Stargazing in the desert during the new moon is really incredible. Hard to believe that we could be alone in the Universe.
We are trying to get these AI cheatsheets accessible to more people around the world.
If you want to help us with the translation into your own native language, you can contribute in this dedicated GitHub repo: github.com/shervinea/cheatsh…
Set of animated Artificial Intelligence cheatsheets covering Stanford's CS 221 class:
Reflex-based: stanford.io/2EqNPHy
States-based: stanford.io/2wh4F7u
Variables-based: stanford.io/2HAiAfh
Logic-based: stanford.io/2M7taia
GitHub repo: github.com/afshinea/stanford…
The table of contents extension is a simple yet very cool add-on that can save you a lot of time when working on a long Jupyter notebook: jupyter-contrib-nbextensions…
Below are some of the best practices on how to make your work reproducible. These are essential when working in teams.
My #AAASmtg talk today on “Collaborating reproducibly” with notes on what I’ll say. biostat.wisc.edu/~kbroman/pr…
This website compiles neatly detailed Python tutorials on a wide range of topics: realpython.com/
Very useful for those who want to get a nice overview of various practical use cases.
Even though technical skills are of course useful and necessary, they are not sufficient. Conveying ideas in a clear and concise manner often proves to be at least as important.
The best way to remember that correlation does not imply causation is to have a funny counterexample in mind. The following website compiles a bunch of them: tylervigen.com/spurious-corr…
All data science projects have parts involving data, analyses and modeling. This website provides a nice overview of some good practices to follow when it comes to structuring your work: drivendata.github.io/cookiec…
Set of illustrated Deep Learning cheatsheets covering the content of Stanford's CS 230 class:
Convolutional Neural Networks: stanford.io/2Rc82Ff
Recurrent Neural Networks: stanford.io/2TO7ur6
Tips and tricks: stanford.io/2PTymHP
PDF version: github.com/afshinea/stanford…
Just found a website that lets you learn SQL in an interactive way. Particularly convenient for those who wish to get some practice. If you're interested, here it is: sqlzoo.net
Additionally, all of the useful SQL commands are summarized in the following doc: github.com/enochtangg/quick-…
