mathstodon.xyz is one of the many independent Mastodon servers you can use to participate in the fediverse.
A Mastodon instance for maths people. We have LaTeX rendering in the web interface!

Server stats:

2.7K
active users

#benchmarking

0 posts0 participants0 posts today

youtu.be/J4qwuCXyAcU

In this video, Ollama vs. LM Studio (GGUF), showing that their performance is quite similar, with LM Studio’s tok/sec output used for consistent benchmarking.

What’s even more impressive? The Mac Studio M3 Ultra pulls under 200W during inference with the Q4 671B R1 model. That’s quite amazing for such performance!

youtu.be- YouTubeEnjoy the videos and music you love, upload original content, and share it all with friends, family, and the world on YouTube.

Surely someone's looked into this: if I wanted to store millions or billions of files on a filesystem, I wouldn't store them in one single subdirectory / folder. I'd split them up into nested folders, so each folder held, say, 100 or 1000 or n files or folders. What's the optimum n for filesystems, for performance or space?
I've idly pondered how to experimentally gather some crude statistics, but it feels like I'm just forgetting to search some obvious keywords.
#BillionFileFS #linux #filesystems #optimization #benchmarking

ZDNet: ‘Humanity’s Last Exam’ benchmark is stumping top AI models – can you do any better?. “On Thursday, Scale AI and the Center for AI Safety (CAIS) released Humanity’s Last Exam (HLE), a new academic benchmark aiming to ‘test the limits of AI knowledge at the frontiers of human expertise,’ Scale AI said in a release. The test consists of 3,000 text and multi-modal questions on more than […]

https://rbfirehose.com/2025/01/28/zdnet-humanitys-last-exam-benchmark-is-stumping-top-ai-models-can-you-do-any-better/

ResearchBuzz: Firehose | Individual posts from ResearchBuzz · ZDNet: ‘Humanity’s Last Exam’ benchmark is stumping top AI models – can you do any better? | ResearchBuzz: Firehose
More from ResearchBuzz: Firehose
Continued thread

5 of these methods can leverage multithreaded (MT) #BLAS with a sweet spot ~ 6 threads for the 40% of the time spent in MT regions. E5-2697 has 36/72 (physical/logical) cores, so the avg case scenario is one in which 0.4x3x6 cores +2 (serial methods) tie up ~ 9.2 cores ~13% of the 72 logical cores. So far the back of envelope calculation, i.e. if I run 5 out of the 2100 design points in parallel, I will stay within 15% of resource use is holding rather well! #benchmarking #hpc #rstats