Run Pytorch in Docker With Cuda Without Pulling Your Hair Out


This week I’ve been working on containerizing some Pytorch based software. Below are some notes on how to make this process more pleasant and relaxing by avoiding a few pitfalls. I used Paperspace Machines, specifically their ML-in-a-Box which is an Ubuntu 20.04 based image that ships with Docker, CUDA and NVidia Docker already installed. Let me point you to this hopefully evergreen SO thread in case your system is missing one or several of these components.…
Read more ⟶

[Notes] A Hackers Guide to LLMs


Notes for Jeremy Howard's video A Hackers' Guide to Language Models
Read more ⟶

Computer Lib Dream Machines


Ted Nelson documented his efforts to create a computer network with a simple user interface via Project Xanadu in 1974-1975. book The Pledge The purpose of computers is human freedom. I am going to help make people free through computers. I will not help the computer priesthood confuse and bully the public. I will endeavor to explain patiently what computer systems really do. I will not give misleading answers to get people off my back, like “Because that’s the way computers work” instead of “Because that’s the way I designed it.…
Read more ⟶

How Smol AI Developer Works


Smol AI/developer is an open source code generation tool by mighty swyx that got a lot of people excited. You can see some demos of its usage here and here. In this post we’ll take a look under the hood to see how it works. After reading this you should be able to recreate something similar potentially better tailored to your needs and your stack. I assume you have some basic experience with LLMs and Python programming language.…
Read more ⟶

Hello World


Hello, world First post to test new setup…
Read more ⟶