AI Linux at Le Robot Hackathon by Hugging Face, hosted at the awesome Miami AI Hub! Robotics needs low-latency edge GPU compute - cloud won’t cut it. Join us to make edge AI effortless. All open source. Let’s build the future 🤖🔥 #AILinux #EdgeAI #Robotics

Leader at work! 🧠💪 Follower is following… mostly in the right direction! #LeRobotWorldwideHackathon LeRobot Hugging Face Miami AI Hub Burhan Sebin

The Real Performance Penalty of GPU Passthrough into a VM? (It's... boring) VM performance was just 1–2% slower than bare metal. That’s it. Practically a rounding error. So… yeah. Turns out GPU passthrough isn’t the scary performance killer. 👉 We put together the full setup,

🚀 We just profiled Google DeepMind Gemma 3n model on an NVIDIA AI GPU with PyTorch Profiler🔬 We’ve open-sourced the code and raw traces so you can reproduce and explore the results yourself. Check it out! 👇 github.com/sbnb-io/gemma3… P. S. Apologies in advance for tagging

Just ran OpenAI's gpt-oss:20b on a budget $429 NVIDIA RTX 5060 Ti 16 GB (Blackwell, 759 AI TOPS) with ollama: 🚀 22/25 layers offloaded into GPU (9.8 GB GPU, 3 GB CPU) 🚀 Prompt eval: 913 tok/s 🚀 Generation: 14.6 tok/s Not bad at all - personal AI at the edge is a

We just submitted to Google’s Gemma3n Hackathon! 🚀 Truly intelligent CCTV that understands live security footage - all locally, no “big brother” control. Cloud would cost thousands/month, but we’re squeezing max performance from an affordable $429 NVIDIA RTX 5060 Ti 16GB

🔥 Incredible - our Reddit post about our AI computer ("GPUter") gained 100K views, 122 comments, and 328 shares! 🚀 reddit.com/r/LocalLLaMA/c… Main takeaways: - Local GPU builds are a hot topic - Commenters mostly recommend RTX 3090 - A few suggestions around more exotic options