AutoGluon-TimeSeries : Creating Powerful Ensemble Forecasts - Complete Tutorial

Amazon's framework for time-series forecasting has it all.

https://aihorizonforecast.substack.com/p/autogluon-timeseries-creating-powerful

How to make LLMs go fast

This post is a long and wide-ranging survey of a bunch of different ways to make LLMs go brrrr, from better hardware utilization to clever decoding tricks.

https://vgel.me/posts/faster-inference

UForm

Pocket-Sized Multimodal AI for content understanding and generation across multilingual texts, images, and video, up to 5x faster than OpenAI CLIP and LLaVA.

https://github.com/unum-cloud/uform

mixtral-offloading

Run Mixtral-8x7B models in Colab or consumer desktops.

https://github.com/dvmazur/mixtral-offloading

falco

Enhance your Django developer experience: CLI and Guides for the Modern Django Developer.

https://github.com/tobi-de/falco

Fine Tuning Python WSGI and ASGI applications for Flask, Django, and FastAPI

In this post, the focus is on examining best practices for configuring Python web servers such as Gunicorn, Uvicorn, and Hypercorn. The architecture of the components between Python code and users will be summarized, and verification methods like load testing will be discussed to ensure the configuration can withstand user traffic.

https://tonybaloney.github.io/posts/fine-tuning-wsgi-and-asgi-applications.html

Fastest Way to Read Excel in Python

Reading 500K rows in less than 4 seconds.

https://hakibenita.com/fast-excel-python

Fontimize: Subset Fonts to Exactly and Only Your Website's Used Characters

Fontimize is a Python library that enables the creation of font subsets containing only the specific glyphs needed for text or HTML, thereby reducing the initial download size of websites and optimizing bandwidth usage.

https://daveon.design/introducing-fontimize-subset-fonts-to-exactly-and-only-your-websites-used-characters.html

MotionCtrl

A Unified and Flexible Motion Controller for Video Generation.

https://github.com/TencentARC/MotionCtrl

semantic-router

Semantic Router is a superfast decision layer for your LLMs and agents. Rather than waiting for slow LLM generations to make tool-use decisions, we use the magic of semantic vector space to make those decisions — routing our requests using semantic meaning.

https://github.com/aurelio-labs/semantic-router

DocFlow

DocFlow is a powerful Document Management API designed to streamline document handling, including seamless uploading, downloading, organization, versioning, sharing, and more.

https://github.com/jiisanda/docflow

Asyncio Logging Without Blocking

You can implement non-blocking logging in asyncio programs by using a shared Queue with a ...

https://superfastpython.com/asyncio-log-blocking/

Retro on Viberary

The article reflects on the author's experience with Viberary, a side project aimed at finding books based on a particular vibe. The project was created to explore machine learning side projects and the intersection of search and recommendations, serving as a production-grade complement to the author's recent deep dive into embeddings.

https://vickiboykis.com/2024/01/05/retro-on-viberary/