The Silicon Trend Tech Bulletin

Icon Collap
...
Home / AI / ML / Nvidia Unites with IBM, Google, and Others for Quantum Research

Nvidia Unites with IBM, Google, and Others for Quantum Research

Published Fri, Nov 12 2021 08:37 am
by The Silicon Trend

nvidia
 

 

Nvidia Unites with IBM, Google, and Others for Quantum Research

During GTC 2021, a US multinational tech firm - Nvidia Corp, revealed 65 updated and new computing libraries, collaborating with other tech giants such as Google, IBM, and more to accelerate quantum research.

 

 

Software Development Kits (SDKs)

Few existing AI-based SDKs that got updated are:

  • Merlin 0.8 - recommendation systems are enhanced to predict the user's next move with little data and support for models bigger than GPU memory.
  • Deepstream 6.0 - computer vision becomes accessible with a visual drag-and-drop interface.
  • Triton 2.15, cuDNN 8.4, and TensorRT 8.2 - helps with the DNN development by offering new optimization for massive language models and inference swiftness.

4 new SDKs are:

  • cuQuantum - designed for swift simulation of larger quantum circuits. Quantum researchers can now simulate fields like error correction algorithms, near-term variational quantum algorithms and speed up popular simulators from IBM, Atos, and Google.
  • ReOpt - automatic optimization of logistical procedures - warehouse selection, fleet combo, and vehicle routes. 
  • CUDA-X accelerated DGL container - assist data researchers and developers in working on graph neural networks for a quick set-up of a working environment.
  • cuNumeric - executes the NumPy application programming interface and allows scaling to multi-node and multi-GPU systems.

 

 

Swift Quantum Research

The first library known as the cuStateVec is from the cuQuantum SDK, Nvidia's initial contribution to partnership with small firms, national labs, university research teams, Google, and IBM. It accelerates the state vector simulation approach that tracks the system's entire memory state, scaling to tens of qubits.

It is incorporated into Google Quantum AI's qsim and can be leveraged via an open-source Cirq. The library will also be combined with a top-notch simulator framework - Qiskit Aer for quantum circuits in December. The national labs that leverage cuQuantum to speed their researchers are MIT, Oak Ridge, Berkeley National Laboratory, Oxford, and more.

The US firm is assisting creators in kicking the start with developing a DGX quantum appliance that places its simulation software optimized for DGX A100 systems.

 

st-3
 

Image Source: https://nvidianews.nvidia.com