Making AMD GPUs competitive for LLM inference Worried about NVIDIA GPU shortage? This project aims to makes it possible to compile LLMs and deploy them on AMD GPUs using ROCm and get competitive performance. Article: https://
blog.mlc.ai/2023/08/09/Mak
ing-AMD-GPUs-competitive-for-LLM-inference
… Discussion: https://
reddit.com/r/MachineLearn
ing/comments/15ml8n0/project_making_amd_gpus_competitive_for_llm/
…
@hardmaru
-
Making AMD GPUs Competitive for LLM Inference with ROCm
By
–
-
Little Book of Deep Learning by François Fleuret Released
By
–
Just received my Little Book of Deep Learning, by @FrancoisFleuret
. At only 150 pages, this will be a fun gem to browse through! The little book is also released under a creative commons license: https://
fleuret.org/francois/lbdl.
html
… -
Understanding: Human Intelligence vs Large Language Models
By
–
Ingredients of Understanding A nice article by @dileeplearning about how human understanding is different from LLM “understanding” https://
dileeplearning.substack.com/p/ingredients-
of-understanding
… -
Breakthroughs in AI Don’t Always Require Moonshot Efforts
By
–
A ‘moonshot’ is usually the last step involved in a breakthrough, and I’d argue that there are many breakthroughs even in AI that didn’t involve or require moonshot-style efforts.
-
Google Brain’s flat structure enabled more impactful research than hierarchical moonshots
By
–
This reminded me of how Google Brain worked when I joined. It was way less hierarchal than DeepMind. Researchers basically self-organized into 3-10 people projects, and given lots of freedom to pursue exploration. This led to more impactful research compared to ‘moonshots’ IMO.
-
Edge Detection Limitations in Understanding Human Vision
By
–
Why Edge Detection Doesn’t Explain Line Drawing “If human vision was just edge detection, then we would have already figured out how vision works a long time ago.” An excellent paper and blog post by @AaronHertzmann from 2021: https://
arxiv.org/abs/2101.09376 -
Running Deep Learning Models on M1 Mac mini
By
–
I wonder if it'll still work if you start running deep learning models or an LLM on this M1 Mac mini.
-
Imperfect approach with simplified tutorial link
By
–
To some extent only. This approach is far from perfect. Here is a (simplified) tutorial from a while back: https://
github.com/hardmaru/slime
volleygym/blob/master/TRAINING.md
… -
Irrational Strategy Against Rational AI Agents Like AlphaGo
By
–
The ‘irrational’ strategy of how to defeat ‘rational’ AI agents reminded me of how Lee Sedol won game 4 vs AlphaGo.
-
Why DeepMind Never Released AlphaGo to the Public
By
–
I wonder this is the reason why DeepMind has never released AlphaGo (or any of its variants), and has never let the general public play against it. So that no one will be able to find holes in it, and show that, just like other AIs, AlphaGo can easily be defeated or jail broken.