iiinova — where your digital self keeps evolving

aymeric-roucher.bsky.social @aymeric-roucher.bsky.social · 31/12/2024

Today we release 𝚜𝚖𝚘𝚕𝚊𝚐𝚎𝚗𝚝𝚜, @hf.co's new barebones library for agentic workflows! 💥 Main logic fits in ~1000 lines of code. 🌍 Supports any LLM through LiteLLM integration. 🛡️ Secure code execution via E2B sandboxes. Try it here 👉 github.com/huggingface/...

170

aymeric-roucher.bsky.social @aymeric-roucher.bsky.social · 04/12/2024

𝗦𝗵𝗼𝘄𝗨𝗜: 𝗮 𝘀𝗺𝗮𝗹𝗹 𝗲𝗻𝗱-𝘁𝗼-𝗲𝗻𝗱 𝗮𝗴𝗲𝗻𝘁 𝘁𝗵𝗮𝘁 𝗰𝗮𝗻 𝗻𝗮𝘃𝗶𝗴𝗮𝘁𝗲 𝗮𝗻𝘆 𝗨𝗜 📲 and beats much larger VLMs! New paper by NUS & Microsoft, agent that acts on any UI (Desktop, Android, Web) without needing additional text information. One great idea: group image patches by GUI group, to speedup and simplify processing.

040

aymeric-roucher.bsky.social @aymeric-roucher.bsky.social · 29/11/2024

Most important thing to do today: go try QwQ on Hugging Chat 👉 huggingface.co/chat/models/...

150

aymeric-roucher.bsky.social @aymeric-roucher.bsky.social · 26/11/2024

The 🐐 Yann LeCun just uploaded MNIST on Hugging Face!

0205

aymeric-roucher.bsky.social @aymeric-roucher.bsky.social · 25/11/2024

🗞️ 𝗦𝘁𝗮𝘁𝗲 𝗼𝗳 𝗘𝗻𝘁𝗲𝗿𝗽𝗿𝗶𝘀𝗲 𝗔𝗜 𝟮𝟬𝟮𝟰: 𝗔𝗻𝘁𝗵𝗿𝗼𝗽𝗶𝗰 𝗲𝗮𝘁𝗶𝗻𝗴 𝘂𝗽 𝗢𝗽𝗲𝗻𝗔𝗜, 𝗔𝗴𝗲𝗻𝘁𝘀 𝗿𝗮𝗺𝗽 𝘂𝗽 𝘁𝗼 𝟭𝟮% 𝗼𝗳 𝘂𝘀𝗲-𝗰𝗮𝘀𝗲𝘀, 𝗼𝗽𝗲𝗻 𝗺𝗼𝗱𝗲𝗹𝘀 𝗺𝗮𝗸𝗲 𝟭𝟵% 𝗼𝗳 𝘂𝘀𝗮𝗴𝗲 Really recommend this report by Menlo ventures! AI spending surged to $13.8 billion this year, more than 6x the $2.3 billion spent in 2023! 👉 menlovc.com/2024-the-sta...

041

aymeric-roucher.bsky.social @aymeric-roucher.bsky.social · 22/11/2024

Made a new app to visualize the LLM race ⇒ 𝗡𝗼 𝗘𝘂𝗿𝗼𝗽𝗲𝗮𝗻 𝗰𝗼𝗺𝗽𝗮𝗻𝘆 𝗶𝗻 𝘁𝗵𝗲 𝘁𝗼𝗽 𝟭𝟬 🇪🇺❌ The top 10 is exclusively US 🇺🇸 and Chinese 🇨🇳 companies (growing fast), with the notable exception of Mistral AI 🇫🇷. See the app here 👉 huggingface.co/spaces/m-ric...

072

aymeric-roucher.bsky.social @aymeric-roucher.bsky.social · 21/11/2024

Lifehack of the day: Adding "r.jina.ai/" before any url transforms it in Markdown using Jina AI's Reader!

020

aymeric-roucher.bsky.social @aymeric-roucher.bsky.social · 20/11/2024

🎓 Training pipeline: ‣ Continued pre-training on Meta's internal docs and wikis ‣ Supervised fine-tuning on past incident investigations ‣ Training data mimicked real-world constraints (2-20 potential changes per incident) Read it in full 👉 www.tryparity.com/blog/how-met...

tryparity.com

How Meta Uses LLMs to Improve Incident Response (and how you can too) - Parity

How Meta Uses LLMs to Improve Incident Response (and how you can too) - Meta used LLMs to root cause incidents with 42% accuracy. Here's how they did it and how you can do it too.

010

aymeric-roucher.bsky.social @aymeric-roucher.bsky.social · 20/11/2024

How did they do it? 🔄 Two-step approach: ‣ Heuristics (code ownership, directory structure, runtime graphs) reduce thousands of potential changes to a manageable set ‣ Fine-tuned Llama 2 7B ranks the most likely culprits

100

aymeric-roucher.bsky.social @aymeric-roucher.bsky.social · 20/11/2024

🤔 42%, isn't that high? ➡️ When there's an issue in prod, engineers dive into recent code changes to find the offending commit. At Meta (thousands of daily changes), this is like finding a needle in a haystack. 💡 So the LLM-based suggestion can cut incident resolution time from hours to seconds!

100

aymeric-roucher.bsky.social @aymeric-roucher.bsky.social · 20/11/2024

🔍 Meta teams use a fine-tuned Llama model to fix production issues in seconds One of Meta's engineering teams shared how they use a fine-tuned small Llama (Llama-2-7B, so not even a very recent model) to identify the root cause of production issues with 42% accuracy.

140

aymeric-roucher.bsky.social @aymeric-roucher.bsky.social · 14/11/2024

The next big social network is not 🦋 It's Hub Posts! [INSERT STONKS MEME LASER EYES] See below: I got 105k impressions since regularly posting Hub Posts, comparable to my 275k on Twitter!

3202