Deci, a deep-learning software maker that utilizes AI-powered tools to help teams create and deploy AI models faster, recently achieved a remarkable breakthrough. All thanks to their exclusive AutoNAC engine for neural architecture construction, a natural language processing (NLP) model generated by Deci’s technology was able to clock 100,000 queries per second in a recent MLPerf Inference v3.0 benchmark. This level of performance makes it the highest ever throughput for NLP-related submissions on MLPerf – 7 times faster than the next-best submission.
To accomplish this feat, Deci used its technology to create architectures tailored to its chosen GPUs – namely the Nvidia A30, A100 80GB, and H100 PCIe GPUs – while keeping the accuracy within 0.1% of the baseline. After testing the models on eight Nvidia A100 GPUs, Deci achieved 103,053 queries per second per TeraFLOPs – 7 times faster than other submissions and offering greater accuracy to boot.
The new benchmark comes shortly after the debut of Deci’s latest version of their deep-learning software. It comes with improved generative AI model optimization as well as a shortcutted time-frame and lower development costs for AI models. Already partnering with businesses like Ibex, Intel, Sight and RingCentral, Deci is well-poised to compete against rivals such as Matlab, Dataloop and DeepCube.
Ran El-Yaniv, Deci’s Chief Scientist and Co-Founder, highlighted the potential that their technology presented, “With Deci’s platform, teams no longer need to compromise either accuracy or inference speed and achieve the optimal balance between these conflicting factors by easily applying Deci’s advanced optimization techniques.” With the success of their latest benchmark, it looks as though Deci is well on its way to realizing its potential as a premier facilitator of AI applications for firms.
Top executives from around the world have now been invited to San Francisco on July 11-12, to learn from industry leaders on how to integrate and optimize investments in AI for maximum success.
Deci looks to save businesses time, money, and resources with its AI models, letting teams get away from the tedious and time-consuming tasks of coding AI applications. Plus, it makes sure that firms are still able to access the performance and accuracy that they need to get the job done. This latest benchmark results only serves to further prove that Deci isn’t just talk.
