GLM-5.2 can now be run locally!π₯
The 2-bit model retains ~82% accuracy after we shrunk it from 1.51TB to 238GB (-84% size).
Run on a 256GB Mac or RAM/VRAM setups.
GLM-5.2 is the strongest open model to date.
Guide: https://t.co/bI7FeeKHDd
GGUF: https://t.co/BMkxswdj5N
I just doubled llama-405b's long context performance on this benchmark:
Same haystack & base prompt, but I added 3 examples that were VERY semantically similar to the real thing.
Result is 70 full accuracy hits, as opposed to the no-shot's 30.
Thanks to @Prof_OZ for the idea
Asking @Google's Gemini and @openai's GPT-4 the exact same prompt tends to yield strangely similar responses which is always fascinating to me. I honestly wonder how this happens. Haven't tried with other large LLMs yet.
@useintro It's hard to admit when a theory you had wasn't 100% right (especially ones in extremely public fields like AGI) but when you have the right people working on what they're good at on a unified goal, things can move fast!
ποΈ Out Now: "Practically Intelligent" E10 - Decoding AI Hype π
Join us and @natolambert as we talk about how to stay accurately informed in AI. ππ€
π Listen for tips on navigating the AI hype cycle! https://t.co/ve219gjURn
#AI#TechPodcast#MachineLearning#Innovation#LLM
Dive into the Exciting World of Large Language Models with Sinan Ozdemir in the new episode of the ODSCβs Ai X Podcast! οΈ
Don't miss this opportunity to gain valuable knowledge from a leading expert @Prof_OZ
Listen now on your favorite podcast platform: https://t.co/tZoviP9Wor
"Get as far as you can without raising a single dollar." - Sophia Amoruso's advice for startup founders:
--
Found this valuable? Repost and follow @useintro to learn how the world's most successful entrepreneurs built $100M+ businesses.
πποΈ "Practically Intelligent" E9 is out! π Join us as we dissect Coatue's AI report and debunk some myths in AI while exploring 2023 trends & what's next in 2024! ππ€
π§ Listen here: https://t.co/QZ1yR1vQkp
#AI#TechTrends#OpenSource#DataScience#Podcast#Innovation
π My new post on #AIAlignment is out today!
π AI vs. Human rating: lots of stark differences
π‘ Key takeaways: scoring mismatch, & positional bias
π Check it out : https://t.co/HcVNB4kbdc
π£οΈ Share your thoughts on #AI's future!
#MachineLearning#DataScience#OReilly
Quick start guides to AI and LLMs, technologies quickly shaping our present and future by @Prof_OZ.
Use direct-to-cart link and save 60%: https://t.co/dIIre0boI8
#ai#llm#datascience#python#gpt
π Shared my insights/resources on Open-Source LLMs at IBM WatsonX's TechXchange today.
Highlights: π₯ Collaboration in open-source π§ Optimizing model efficiency π Importance of quality data & labeling in AI
π Slides & Colab: https://t.co/clv1hBKbEw
#LLMs#AI#ML
π#PracticallyIntelligent Ep 7 is out! π§ Join us and @heuristicity for a deep dive into #AI Benchmarking. We explore benchmarks in AI progress π and the fine line between conceptual & rote learning in AI π
π Listen now: https://t.co/Rt5HMCedgq
#MachineLearning#LLMs