Damn we gotta chat! I was a nuclear engineer before 🙂
COMPUTING
-
Evaluating Resource Consumption and Value Proportionality Beyond Size
By
–
Yes I did, and I think it’s a useful comparison. I also think it’s good to evaluate resource consumption beyond size comparisons and consider whether the value it brings is proportional
-
Choosing Mac Hardware for Local LLM and Vector Database Setup
By
–
depends. if you just want to surf and blog, just get an air. If you want a Miqu-1 Q4 (or Q5 with a teensy ctx window) attached to a chroma db locally, I'd say get a Max with some beefy memory.
-
GPU Memory Requirements for Document Embedding Storage
By
–
I'd guess pretty poorly. but you really don't need that much GPU memory to store a fairly large corpus. all you need is the bag-of-words information for each document; that's the "embedding". eventually you will run out though, and I don't have any hierarchical logic built in.
-
MacBook Pro Battery Life Achievement During Extended Work
By
–
AI is cool and all but so far the most underrated thing about the future is that I've been working from bed while sick for like three days and I have yet to have to charge my macbook pro.
-
BM25 Search Implementation Performance Comparison with ElasticSearch
By
–
also shout out to @clifapt for running BM25 via ES on LoCo first: https://
x.com/clifapt/status
/1755303087873429667
… my implementation seems to be slightly less performant than ElasticSearch; not sure why. maybe because i'm using a subword tokenizer, or suboptimal hyperparams? -
Fast GPU-Enabled BM25 Implementation in PyTorch Achieves SOTA
By
–
implemented a fast, GPU-enabled BM25 in pytorch! BM25 is a simple search algorithm from the 70s that works as well as neural networks for most search problems; for all the advances we've made in neural text retrieval, it's still around got near SOTA on stanford LoCO benchmark
-
Data Compression and Upload Optimization for AI Systems
By
–
Simple, upload everyone and use data compression to collapse the npcs
-
S3 Bucket Analytics: Data Security Risks Without Sanitization
By
–
Analytics with naive chunking except instead of periods it’s entire S3 buckets. Table sanitization is for losers.
-
Selene and Eos Supercomputers Power Next Generation Generative AI
By
–
At #GTC24, hear from the architects of Selene and Eos about how they designed these supercomputers. Learn more about the next generation of #NVIDIADGX architecture to power your #generativeAI applications. #DataCenter
— NVIDIA AI (@NVIDIAAI) 29 février 2024
Register now. https://t.co/ZP6Y95d8Ve pic.twitter.com/O9dzPKFmcBAt #GTC24, hear from the architects of Selene and Eos about how they designed these supercomputers. Learn more about the next generation of #NVIDIADGX architecture to power your #generativeAI applications. #DataCenter Register now. https://
nvda.ws/48QWgYh