What happens if we remove 50 percent of Llama?
Discover Sparse Llama: A 50% pruned, GPU-optimized Llama 3.1 model with 2:4 sparsity, enabling faster, cost-effective inference without sacrificing accuracy.
The latest blog posts to take off on Hacker News. 5,273 posts.
Discover Sparse Llama: A 50% pruned, GPU-optimized Llama 3.1 model with 2:4 sparsity, enabling faster, cost-effective inference without sacrificing accuracy.
It was not easy or quick, and there were so many qualifiers that, for over a decade, I didn’t really think I “deserved” the label. But as I got older, those qualifiers started dropping off one by one, and today I’m convinced that I have perfect pitch. I describe my musical journey and thoughts on what perfect pitch…
The Netscapes of AI
Virtualization is a powerful technique for code obfuscation, and reversing it can be challenging. In this post, we cover the work done during an internship on developing an automated devirtualization tool. We explore a simplified taint-based approach and discuss its limitations. For a more in-depth analysis, the full…
Dmitrii Kovanikov writes posts about enterprise software development and functional programming, both entertaining and serious. I never used a functional programming language (such as Haskell or OCaml) at work, but several ideas from functional programming have become popular in mainstream general-purpose programming…
Kenji López-Alt posed a mathematical problem on his YouTube channel about how to best cut an onion. I present a detailed solution.
On the last day of 1945, with World War II finally behind it, Finland's government announced a new and very strange policy. All Finns were r...
Discover more about what's new at AWS with Amazon S3 adds new functionality for conditional writes
[“The welcoming, friendly and visually pleasing appearance” of the TSA’s headquarters at 6595 Springfield Center Drive, Springfield, VA.]
A guide to the operating principles, techniques and technology in lidar systems for self driving cars.
Artist in Residence Peter Hristoff meditates on the role of Islamic prayer rugs (seccades) and the role they play in his own creative output.
Oracle is holding the JavaScript trademark hostage, and we’re pursuing legal means to #FreeJavaScript. Here’s a brief update.
A deep dive into the complexities of optimizing code for SIMD instruction sets across multiple platforms.
MCP is an open standard for connecting AI assistants to the systems where data lives, from content repositories to business tools.
A data scientist’s perspective
I built a new plugin for my sqlite-utils CLI tool that lets you ask human-language questions directly of SQLite databases and CSV/JSON files on your computer. It’s called sqlite-utils-ask. Here’s …
And the content website wasteland
Institutional memory comes in two forms: people and documentation. People remember how things work and why. Sometimes they write it down and store that information somewhere. Institutional amnesia works similarly. The people leave and the documents disappear, rot, or just become forgotten (as it were).
The Royal Society is a venerable institution founded in 1660, whose original members included such eminent men as Christopher Wren, Robert H...
In 2022, three companies with strange names and no clear business purpose beyond patent litigation filed dozens of lawsuits in Delaware federal court, accusing businesses of all sizes of patent infringement. Some of these complaints claimed patent rights over basic aspects of modern life; one, for...
Why are database DELETEs complex and inefficient? Uncover challenges from physical data removal to autovacuum, and find strategies for managing large deletions.
Update: This article was discussed on Hackernews Much has been written and a lot of analysis performed on the global BGP table over the years, a significant portion by the inimitable Geoff Huston.
This may be overkill, but it works on my machine
The Epoch Cassette Vision was a moderate success. But in 1983, that all ended, when Nintendo and Sega released new consoles, which had more advanced hardware...
This is directly inspired by Engineering for Slow Internet. I figured I’d give running with dogshit internet on my desktop and phone a go to see how poorly (or well!) specific applications behave. This is not an in-depth review of the behavior, but just my general impressions running for a few hours on 200kbit…
Camera traps and drones deployed by government authorities to monitor a forest in India are infringing on the privacy and rights of local women.
Date: March 7, 2111
The dream of a single dialect-free C++ has probably been dead for many years, anyway.
Well, not this one. But this one is! How? Let’s take a closer look at Bluesky and the AT Protocol that underpins it. Note: I communicated with the Bluesky team prior to the publishing of this post. While the functionality described is not the intended use of the application, it is known behavior and does not constitue…