Deep learning compilers and how the sausage is made

A series of discussions focused on deep learning compilers and machine learning. It covers how deep learning models are optimized and deployed for practical use cases. The content dives into topics relevant for developers and engineers working with AI frameworks and model performance optimization.

https://www.youtube.com/playlist?list=PLaMu-SDt_RB7ImARcTT_Wjypwx2vBIBen

MiniLang

A type-safe C successor that compiles directly to various platforms.

https://github.com/NICUP14/MiniLang

WordLlama

Things you can do with the token embeddings of an LLM.

https://github.com/dleemiller/WordLlama

Some tricks with UV

UV can be seen as an alternative to pip, but that might be a limiting way to think about the tool. Instead of looking at faster builds, which are still super nice, it might also make sense to rethink the stuff that we might be able to do from Python going forward.

https://www.youtube.com/watch?v=jXWIxk2brfk

ft_utils

A library of utilities to support efficient, scalable Python development leveraging Free Threaded Python.

https://github.com/facebookincubator/ft_utils

FARM Stack Course – Full Stack Development with FastAPI, React MongoDB

Learn full stack stack development with the FARM stack. You will learn to quickly create an application using FastAPI, React, and MongoDB.

https://www.youtube.com/watch?v=PWG7NlUDVaA

Formatron

Formatron empowers everyone to control the format of language models' output with minimal overhead.

https://github.com/Dan-wanna-M/formatron

RAG Is More Than Just Vector Search

Go beyond vector search. Learn how to improve your RAG system with Text2SQL, filtered search, structured extraction, and eval-driven development.

https://www.timescale.com/blog/rag-is-more-than-just-vector-search/

What's in an e-graph?

The article explains e-graphs by incrementally building from union-find to a full e-graph implementation, highlighting key features like equivalence class discovery, pattern matching, and extraction. It demonstrates how e-graphs can be used in compilers for optimizations, offering a more flexible alternative to traditional find-and-replace methods while discussing trade-offs and variatio...

https://bernsteinbear.com/blog/whats-in-an-egraph/

spann3r

3D Reconstruction with Spatial Memory.

https://github.com/HengyiWang/spann3r

Serializing package requirements in marimo notebooks

Marimo now allows notebooks to serialize their package requirements as top-level comments, enabling users to run notebooks in isolated virtual environments with a single command. This feature, powered by the uv package manager, enhances reproducibility and sharing of notebooks by eliminating the need for separate requirements files and preventing environment pollution.

https://marimo.io/blog/sandboxed-notebooks

Let’s build and optimize a Rust extension for Python

Python code too slow? You can quickly create a Rust extension to speed it up.

https://pythonspeed.com/articles/intro-rust-python-extensions/

LLaMA-Omni

LLaMA-Omni is a low-latency and high-quality end-to-end speech interaction model built upon Llama-3.1-8B-Instruct, aiming to achieve speech capabilities at the GPT-4o level.

https://github.com/ictnlp/LLaMA-Omni

FastAgency

The fastest way to bring multi-agent workflows to production.

https://github.com/airtai/fastagency

Deploying a Django app with Kamal, AWS ECR, and Github Actions

The article provides a comprehensive guide on deploying a Django app using Kamal, AWS ECR, and GitHub Actions, offering a streamlined approach to containerized deployment. It covers setting up a VPS, creating a Dockerfile, configuring AWS ECR, setting up Kamal, and automating the deployment process with GitHub Actions, aiming to simplify the deployment workflow for developers.

https://dylancastillo.co/posts/deploy-a-django-app-with-kamal-aws-ecr-and-github-actions.html

pyrtls

rustls-based modern TLS for Python.

https://github.com/djc/pyrtls