[ad_1]
Meta and NVIDIA have introduced that they’re engaged on constructing a large AI supercomputer. The AI Analysis SuperCluster (RSC) is presently coaching new fashions to advance AI.
The RSC is anticipated to be the biggest buyer set up of NVIDIA DGX A100 techniques as soon as it’s totally deployed. It’s anticipated to be totally constructed out later this 12 months, and it will likely be used to coach AI fashions with over one trillion parameters. The supercomputer will probably be helpful for a lot of completely different fields like natural-language processing (NLP).
Based on Meta, the corporate is concentrated on enabling efficiency at scale and excessive reliability, safety, privateness, and the flexibleness to deal with varied AI fashions.
Specifics of the RSC
The RSC makes use of 760 NVIDIA DGX A100 techniques as its compute nodes, they usually have 6,080 NVIDIA A100 GPUs linked on an NVIDIA Quantum 200Gb/s InfiniBand community. All of this permits it to ship 1,895 petaflops of TF32 efficiency.
The RSC solely took 18 months to succeed in a working AI supercomputer, which is spectacular given the COVID-19 pandemic’s impact on growth.
Early benchmarks on the RSC, in contrast with Meta’s legacy manufacturing and analysis infrastructure, have demonstrated that it might probably run laptop imaginative and prescient workflows as much as 20 instances sooner. It could possibly additionally run the NVIDIA Collective Communication Library (NCCL) over 9 instances sooner and prepare large-scale NLP fashions 3 times sooner. In different phrases, a mannequin with tens of billions of parameters would have the ability to end coaching in three weeks, in comparison with the earlier nine-week benchmark.
What the RSC Will Allow
The RSC will allow Meta AI researchers to create new AI fashions that may study from trillions of examples. Researchers will even have the ability to work throughout lots of of various languages; analyze textual content, photos, and video collectively; develop new augmented actuality instruments; and extra.
Meta hopes that the RSC will allow the corporate to energy real-time voice translations to massive teams of individuals, even when every is talking a unique language. It will permit extraordinarily numerous groups to collaborate on analysis tasks.
“We hope RSC will assist us construct fully new AI techniques that may, for instance, energy real-time voice translations to massive teams of individuals, every talking a unique language, so they may seamlessly collaborate on a analysis venture or play an AR sport collectively,” Meta stated.
This new collaboration will go a good distance in growing one of many subsequent state-of-the-art AI supercomputers that can be utilized in a variety of industries.
[ad_2]
