AI HARDWARE

Groq LPU System Achieves 240 Tokens Per Second with Llama-2

By

–

31 August 2023 16h09

Our LPU™ system is pushing the limits on LLM #inference perf again, now running Llama-2 70B at 240 tokens per sec per user! CEO @JonathanRoss321 shares more on the >2x improvement, why ultra-low latency matters, and if GPUs can still catch up. More at http://
groq.link/240tps

Groq Platinum Sponsor at AI for Defense Summit 2023

By

–

31 August 2023 16h01

We are proud to be a Platinum Sponsor at this year's AI for Defense Summit in National Harbor, MD. Stop by to see a demo of how Groq is enabling best-in-class performance with our software-defined deterministic chip, the Language Processing Unit™ (LPU™)

→ View original post on X — @axeleraai

Generative AI at the Edge: Guest Lecture at Taiwan-Eindhoven

By

@axeleraai

–

31 August 2023 14h00

Join our very own Bram-Ernst Verhoef tomorrow for his guest lecture "Moving Generative AI Towards the Edge" at the inaugural Summer School Taiwan-Eindhoven, hosted at @TUeindhoven
. https://
etspss.nl/home

Groq Day: AI Acceleration and LLM Inference Solutions

By

–

31 August 2023 1h00

Join us at #GroqDay to understand why now is the moment that matters for AI acceleration of LLMs and how Groq inference solutions provide customer advantages. Register at http://
groq.link/groqday5

→ View original post on X — @esyudkowsky

Image Models Limitations: Computational Constraints and Reduced Capability

By

@esyudkowsky

–

30 August 2023 20h12

Images are hundreds of thousands of pixels, so nobody can afford an architecture that runs a large model over every pixel. The resulting small models are in fact much stupider than GPT-4 and have trouble following even slightly complicated instructions.

→ View original post on X — @jeffdean

Montreal Datacenter: Hardware Mix and CO2 Impact Timeline

By

@jeffdean

–

30 August 2023 17h17

I think Montreal is a relatively new datacenter (but great for CO2e!). The land purchase was just announced in May, 2021, so with construction and equipment deployment timelines, it may not be fully populated, so mix of hardware there may change.

→ View original post on X — @jeffdean

Google Cloud Datacenters Carbon Intensity and TPUv4 Sustainability

By

@jeffdean

–

30 August 2023 16h15

Depends on the data center (I actually don't know which one this was filmed in). We report %age of Google carbon free energy as well as Grid carbon intensity
(gCO2eq/kWh) per Cloud datacenter here. Most TPUv4 pods are installed in ones marked "Low CO2" https://
cloud.google.com/sustainability
/region-carbon
…

Groq showcases record-breaking LPU inference performance at ORNL conference

By

–

30 August 2023 16h04

Stop by our table at @ORNL
's Smoky Mountains Computational Sciences & Engineering Conference to to get a sneak peak at the record breaking inference performance on the @GroqInc Language Processing Unit™ (LPU™) system, which we'll be announcing publicly tomorrow.

→ View original post on X — @cerebras

Jais: Advanced Arabic LLM Trained on Condor Galaxy 1

By

@cerebras

–

30 August 2023 15h24

(3/3) Jais was trained on the newly unveiled Condor Galaxy 1 (CG-1) AI supercomputer, built by the G42 – Cerebras strategic partnership. To learn more, check out the Press Release: https://
cerebras.net/press-release/
meet-jais-the-worlds-most-advanced-arabic-large-language-model-open-sourced-by-g42s-inception
…