(3/4) This work exploits the high memory bandwidth of the Cerebras CS-2 for seismic processing by leveraging low-rank matrix approximation to fit the problem on SRAM and using many wave-equation-based algorithms that rely on multidimensional convolution operators.
AI HARDWARE
-
Gordon Bell Prize Recognizes State-of-the-Art Scientific Computing
By
–
(4/4) The Gordon Bell Prize will be announced at #SC23 and is awarded to the most valuable scientific computation as demonstrated using state-of-the-art software and hardware technologies on world-leading supercomputers. We're honored to be considered for this prestigious award!
-
KAUST collaborative research finalist for 2023 Gordon Bell Prize
By
–
(1/4) We are thrilled that our collaborative research with the King Abdullah University of Science and Technology is among the finalists for the 2023 Gordon Bell Prize! https://
sc23.supercomputing.org/2023/08/a-look
-at-the-2023-gordon-bell-prize-finalists/
… -
Cerebras CS-2 Scales Memory Wall for Seismic Data Processing
By
–
(2/4) The nominated work is titled, "Scaling the 'Memory Wall' for Multi-dimensional Seismic Processing with Algebraic Compression on Cerebras CS-2 Systems." This work uses Condor Galaxy 1, the AI Supercomputer we built in collaboration with G42. https://
cerebras.net/condor-galaxy-1 -
Enterprise LLM Inference Deployment at Scale
By
–
Real-time, highly accurate insights, at a price point supportive of business needs at scale, are critical to evolving markets. For guidance, check out @aeaglejr
's white paper, Key Enterprise Considerations for Inference Deployment of Large Language Models. http://
groq.com/inference/ -
Optogenetic setups becoming mainstream in AI research
By
–
who isn't running an optogenetic setup nowadays
-
Groq Launches Ultra-Low Latency Generative AI with Llama-2
By
–
Ultra-low latency #generativeAI by @GroqInc is here. Schedule your private demo viewing of Llama-2 70B running on a Groq LPU™ by reaching out to contact@groq.com.
-
Groq Achieves 100 Tokens Per Second Performance on Llama2
By
–
100 Tokens per second per user on #Llama2 from @MetaAI! This ultra-low latency performance could have a massive impact on workloads using #LLMs for everyone from artists to analysts, programmers to educators, all #GenAI and beyond. Book your demo to learn more: contact@groq.com pic.twitter.com/VWcSqPRm18
— Groq Inc (@GroqInc) 8 août 2023100 Tokens per second per user on #Llama2 from @MetaAI
! This ultra-low latency performance could have a massive impact on workloads using #LLMs for everyone from artists to analysts, programmers to educators, all #GenAI and beyond. Book your demo to learn more: contact@groq.com -
Groq Achieves 100 Tokens Per Second With Llama-2 70B
By
–
Announcement. @GroqInc is the first to accomplish 100 tokens per second, per user, running @MetaAI Llama-2 at 70B parameter size as an #LLM . No kernels or CUDA libraries necessary! Save thousands of developer hours with our deterministic Compiler methods.
-
NVIDIA Announces GH200 Superchip and AI Workbench Updates
By
–
Recap highlights from our special address at #SIGGRAPH2023, including the updated GH200 Grace Hopper Superchip, NVIDIA AI Workbench, and updates on @NVIDIAOmniverse with generative #AI. https://
nvda.ws/450MVM9