Zapata’s Orquestra targets hybrid quantum-classical challenge

As it has become clearer that hybrid quantum-classical computing solutions are likely to be needed to achieve practical quantum computing, there has been increasing emphasis on developing software platforms to build applications and hybrid workflows. The recently announced QODA (Quantum Optimized Device Architecture) by Nvidia is among the notable new features. Zapata Computing is an early pioneer of QC software that has banked on leveraging hybrid quantum applications using its Orquestra platform to deliver a quantum advantage.

Timothy Hirzel, Chief Evangelist for orchestrarecently provided HPCwire with an update on progress on the heels of Nvidia’s QODA announcement, which named Zapata as a former collaborator. (See HPCwire cover, Nvidia dives deeper into Quantum, announces programming platform QODA.)

Nvidia’s intention, of course, is to integrate its GPU accelerator technology with quantum systems for use that takes advantage of hybrid classical-quantum resources. Zapata has a bigger goal. Its Orquestra product integrates with tools, such as QODA (when available), to build and deploy applications that run on a wider variety of hybrid quantum-classical resources. It’s not a new idea. Agnostiq’s Covalent platform, for example, is similar but intended for the R&D environment for prototyping and testing. Orquestra, says Hirzel, is aimed squarely at enterprise deployment and production environments.

Do not mistake yourself. No one is using Orquestra as a production deployment tool yet, but one collaborator – Andretti Autosport – is building an infrastructure that includes plans to use Orquestra in production. The motor racing powerhouse intends to use Orquestra and blended quantum-classical resources for a variety of applications such as tire degradation models and racing simulations to understand the likelihood of a yellow flag or crash. a warning ahead.

“Some of these models need to be run trackside and in real time,” Hirzel said. “Andretti has been a fantastic forcing feature on the software because things that weren’t as critical to an R&D user are now very important.”

Most of Zapata’s early customers are still working in R&D, building quantum roadmaps and determining which high-value problems can benefit from quantum resources. That said, the emergence of platforms, such as Orquestra and Covalent and QODA, to develop hybrid applications and manage hybrid quantum-classical resources is an important trend. The market is still going and worth watching. Increasingly, it looks like quantum computing will require a mixed classical-quantum environment.

Hirzel said, “We see quantum and classical growing together. By no means do we see quantum actually replacing HPC, or classical. As quantum devices grow, the number of qubits increases, and the gate depth increases, the challenge of compiling circuits will increase with it. In a simple experiment, you could run millions of shots on a quantum device. You’re going to have to use a substantial amount of classical processing alongside quantum, not just for compiling circuits, but for all data pre-processing and post-processing.

In addition to these necessary processing tasks, a lot of work is underway to understand which parts of applications (optimization, simulations, etc.) are better handled by classical resources and which by quantum. Zapata has long been a champion of this line of thinking. For example, Zapata has worked extensively with quantum enhanced solvers using machine learning and quantum generated random numbers to improve performance.

CEO Christopher Savoie said HPCwire one year ago:

  • “All machine learning basically starts with a generative model, at least in modeling things, a probability distribution made up of a string of random bits. You start with an adjustable, random string of bits to get closer and closer – like a Born machine (a machine born from a quantum circuit, QCBM) – to model the distribution that will give you a certain result. It is your priority. [Next], you power a distro and hopefully get a better handwriting sample, better portfolio, better optimization of something, chemistry, whatever, in the end. This is the result. [To do that] you need a random bit string that has an adjustable parameterized connection to it. The distribution of these bit strings is the source of your model’s wealth. That’s the premise.
  • “What we have shown in our papers is that if you have a quantum source of bitstrings, you get better distribution and better handwriting sample and better portfolio optimization than you can with classical machine work. We can do it today. What we’ve done is we’ve taken a quantum enhanced neural network approach to basically put our quantum spy on this solver. We power a classic neural network workflow that already exists and is already in production in many places. What’s great is that their cost function, their model, already exists too; we just put a little quantum spy on it. We think it’s pretty smart, and it’s patented, and very disruptive. We are able to model the distribution of correct answers for any software.

Using quantum-generated random numbers as input to improve classical solvers is just one example and probably represents a low-hanging fruit in the development of classical-quantum hybrid solutions. But the point is that there will be more complicated problems, some of which are better suited to classical or quantum computing. Nvidia already argues that matrix multiplications, for example, are handled better on GPUs than on QPUs.

Platforms to develop these quantum enhanced applications and workflows will be needed. Enter Orquestra, QODA, Covalent. There will be more such offerings, covering various parts of the required hybrid stack, if you will. Zapata thinks he has an advantage because he has always focused on hybrid approaches.

Looking at the figure above, the solid green areas represent proprietary Zapata technology, the light green combines Zapata and user-provided technology, and the gray areas are non-Zapata resources, hardware, and software. with which Orquestra interacts. The Orquestra stack works with a variety of popular quantum languages ​​and libraries as well as classical simulators and quantum hardware. The idea is to provide a widely usable platform by leveraging existing and new tools.

Hirzel notes that the number of qubit types available (ion trap, superconductor, etc.) is a moving target and each requires its own lower-level protocols and tools. Although the number of qubit technologies is currently limited to a handful, this could change. Zapata supports the major qubit technologies currently in use and plans to expand support as needed.

A central question that hybrid development platforms should help answer, Hirzel said, is which computational engine (classic or particular qubit technology) is better, and why, for a given application, including which part of the app. This is in addition to the development and execution of the workflow.

“Deciding where you want to send your compute work becomes its own challenge. In a more limited way, we still see it today with how organizations are learning to take advantage of GPUs. It’s really there that we want Orquestra to live,” Hirzel said. “Looking back, it’s not just where things are computed. You also have to think about where the data is stored; what kind of orchestration layer do you have sitting on it? slurm at HPC? Does it work on the cloud? And the possibility of visualizing the data coming out of it. »

The platform is still in its infancy and users tend to bring a fair amount of expertise. Quantum-Ready Application functionality, as described by Hirzel, is currently more about taking an existing application and using the Orquestra tools to prepare it to run on various resources, including quantum or classical (e.g. GPU), and then run comparison simulations.

Hirzel said, “It’s about building useful tools and that’s something we potentially see our own internal service teams doing. But ultimately, partners can take it upon themselves to write apps themselves and use Orquestra to deliver apps. It is not a toolkit for creating GUIs and the like, but rather for creating the services that support existing user applications. An end user may have an existing tool – a computational fluid dynamics tool used for aircraft wing design – and is really looking to speed up some of that tool.

He says the user can use Orquestra tools to break down the application and simulate various approaches to determine effectiveness among various device types. The quantum version(s) could then be deployed from Orquestra.

Zapata is also working on assembling algorithm suites for various functions and devices. For example, “Our hardware team is working on algorithms to help users select the most appropriate device for this circuit. Ultimately, we would be looking to put this into something like a quantum hardware broker. Users could then provide selection criteria to the broker – availability, cost, loyalty, speed, etc. – and this one would offer options.

Don’t underestimate the value of being able to estimate the costs of running a quantum workflow, Hirzel said, “We’ve noticed in our time running these quantum simulations or running on real quantum devices, that it’s very easy to run a big bill quickly. This is a problem a lot of people have in the cloud or in HPC.” Provide a cost comparison for various calculation methods during the simulation or just before “we consider as actually quite integral and can be a nice feature in Orquestra”.

Orquestra is an early entry into the hybrid classical-quantum software development platform market. It is still evolving and so is the market segment. Zapata, who graduated from Harvard in 2017, has grown rapidly. The membership is already over 100. He hopes his pioneering status and deep technical expertise will translate into success. Again, this sounds like many start-ups in the quantum computing landscape. Stay tuned.

Comments are closed.