Amazon Braket FAQs

The Amazon Braket Software Development Kit (SDK) is a technology-agnostic developer framework that allows you to develop quantum algorithms and run them on different quantum computing hardware and simulators through the Amazon Braket service. The SDK helps you track and monitor quantum tasks submitted to Amazon Braket and evaluate the results. The Amazon Braket SDK includes a local quantum circuit simulator that you can use to test your algorithms.

Amazon Braket provides fully managed Jupyter notebooks that come pre-installed with the Amazon Braket SDK and example tutorials that help you get started quickly. The Amazon Braket SDK is open sourced so you can use Amazon Braket from any local integrated development environment (IDE) of your choice.

Today’s quantum computers are noisy, and researchers often need access to the lowest level of control over the hardware in order to study noise or cross-talk, develop new and more robust gates, devise error mitigation schemes, and explore novel quantum algorithms. With pulse-control, you can manipulate the low-level, analog signals or pulses that control the qubits of a quantum processor.  

Yes. You can program quantum circuits on Rigetti Computing and Oxford Quantum Circuits quantum computers directly using gates, pulses, or a combination thereof. You can also selectively insert blocks of pulse instructions in specific areas of the program to focus on optimizing individual operations and fine tune performance. 

PennyLane is an open source software library for variational quantum computing that integrates with Amazon Braket. Variational quantum computing is a paradigm that utilizes hybrid quantum-classical algorithms to iteratively find solutions to computational problems in a variety of domains, such as chemistry, optimization, and quantum machine learning. Built around the concept of quantum differentiable programming, PennyLane allows you to train quantum circuits the same way as neural networks. It provides interfaces to popular machine learning libraries, including PyTorch and TensorFlow, to make training your quantum algorithms easy and intuitive. You can learn more about PennyLane at https://pennylane.ai , and read our developer guide here . 

Near-term quantum computing applications in chemistry, optimization and quantum machine learning are based on variational quantum algorithms that utilize iterative processing between classical and quantum computers. PennyLane makes it easy to get started and build variational and quantum machine learning algorithms on Amazon Braket. It allows you to use familiar tools from machine learning to build and train your algorithms. PennyLane provides a chemistry library, qchem, which you can use to map a computational chemistry problem to a quantum computing formulation with a few lines of code.

Amazon Braket helps you to innovate faster with PennyLane. When testing and fine-tuning your algorithms, our fully managed, high-performance on-demand simulators speed up training by 10x or more, compared to simulating your algorithms locally.  To speed up your hybrid quantum algorithms, you can now leverage high performance embedded simulators from PennyLane, such as the lightning.gpu simulator accelerated by NVIDIA’s cuQuantum SDK for GPU based workloads. These simulators come with methods such as the adjoint method for gradient computation which reduce the number of circuits required to compute and gradient and can be used for fast iterative experimentation and prototyping. 

Amazon Braket notebooks come pre-configured with PennyLane, and our tutorial notebooks help you get started quickly. Alternatively, you can install the Amazon Braket PennyLane plugin for any IDE of your choice. The plugin is open source and can be downloaded from here . You can find the PennyLane documentation at https://pennylane.ai .  

OpenQASM is an open source intermediate representation (IR) for quantum computing programs. You can run OpenQASM programs on all gate-based Braket devices either through the Amazon Braket SDK or by directly submitting them to the Braket API. AWS has joined the OpenQASM steering council to help build an open, hardware-agnostic, and unified specification for gate-based quantum programs. 

Error mitigation encompasses a variety of methods to reduce the effects of systematic noise on today’s error-prone hardware by mapping an input circuit to a set of related circuits and applying classical post-processing to the results.

Yes. You can experiment with error mitigation on the IonQ Aria QPUs using IonQ's debiasing technique. Note that this type of error mitigation requires using a minimum of 2,500 shots per task.

Quantum circuit simulators run on classical computers. With simulators, you can test your quantum algorithms at a lower cost than using quantum hardware, and without having to wait to access specific quantum machines. Simulation is a convenient way to quickly debug quantum circuits and to troubleshoot and optimize algorithms before progressing to run them on quantum hardware. Classical simulation is also essential to verify the results of near-term quantum computing hardware and study the effects of noise.

Amazon Braket offers you a choice of four quantum circuit simulators: the local simulator in the SDK, and three on-demand simulators: SV1, a general purpose quantum circuit simulator, DM1 that enables you to simulate the effect of noise on your circuits, and TN1, a high performance tensor network simulator. With these options, you can choose the approach that best suits your requirements.

The local simulator is included in the Amazon Braket SDK at no cost. It can run on your laptop, or within an Amazon Braket managed notebook. You can use it for quick validation of circuit designs. It is well suited for small and medium-scale simulations – up to 25 qubits without noise, or up to 12 qubits with noise, depending on your hardware.

SV1 is a fully managed, high-performance state vector simulator for quantum circuits up to 34 qubits. As a state vector simulator, it takes the full wave function of the quantum state and applies the operations of the circuit to calculate the result. After you have designed and debugged your quantum algorithm using the local simulator in the Amazon Braket SDK, you can use SV1 for scaled testing and research. SV1 automatically scales classical compute resources so you can run up to 35 simulations in parallel.

DM1 is a fully managed, density matrix simulator that enables you to investigate the effects of realistic noise on your quantum algorithms. This can help you develop error-mitigation strategies to get more accurate results from today’s quantum computing devices.

DM1 supports the simulation of circuits up to 17 qubits. It can run up to 35 simulations in parallel, to speed up your experiments. For rapid prototyping and debugging before using DM1, you can use the local noise simulator in the Amazon Braket SDK.

TN1 is a fully managed, high-performance tensor network simulator that is used for structured quantum circuits up to 50 qubits in size. A tensor network simulator encodes quantum circuits into a structured graph to find the best way to compute the outcome of the circuit. TN1 is well suited for simulation of sparse circuits, circuits with local gates, and circuits with inherent structure.

SV1 is a general purpose simulator based on state vector technology. It provides predictable execution and high performance for universal circuits up to 34 qubits.

DM1 is specifically designed to support noise modeling. If you need to study your algorithms under the effects of various types of noise, use DM1.

TN1 is a specialized simulator for certain types of quantum circuits with up to 50 qubits. Consider it for sparse circuits, circuits with local gates and other circuits with inherent structure. Other circuit types, such as those with all-to-all connectivity between qubits, are often better suited to SV1.

Today’s quantum devices are inherently noisy. Every executed operation has a chance of introducing an error. Consequently, the results obtained from a quantum computer generally differ from what is ideally expected. DM1 enables you to study the robustness of your algorithms under the effects of realistic noise, and build error mitigation strategies that help to get more accurate results with today’s quantum computing devices.

DM1 can simulate circuits without noise. However, for best performance, we recommend using SV1 for large-scale simulations of noise-free circuits.

Not if you are using an Amazon Braket on-demand simulator. When using SV1, TN1, or DM1, Amazon Braket manages the software and infrastructure for you. You just need to provide the circuit to run.

If you are running the local simulator in the SDK on your Amazon Braket managed notebook, it will run on the Amazon instance you have already specified for your notebook.

As long as your circuit is within the qubit number and circuit depth limits described here , TN1 will attempt to simulate it. In contrast to SV1, however, it is not possible to give an accurate estimate of the runtime based on the qubit number and circuit depth alone. During the so-called “rehearsal phase”, TN1 will first try to identify an efficient computational path for your circuit, and estimate the runtime of the next stage, the “contraction phase”. If the estimated contraction time exceeds the TN1 limit, TN1 will not attempt contraction and you pay only for the time spent in the rehearsal phase. To learn more, visit the technical documentation .

No, with Amazon Braket, you can direct the same quantum circuit to run on any simulators and gate-based quantum hardware available on the service by changing a few lines of code.

On the AWS Marketplace you can find offerings that use the state-of-the-art classical approaches to combinatorial optimization problems such as Meta Analytics and the Toshiba SBM. 

Running your circuit design on an actual quantum processing unit (QPU) is easy. Once you have created your circuit or problem graph in the Amazon Braket SDK, you can submit your task from within a managed Jupyter notebook or any IDE of your choice, such as PyCharm.

The steps to run a quantum task on a QPU are the same as running on a simulator, you simply chose the back-end, or device, when making API calls within the Amazon Braket SDK. They are both compute operations for which you can request different back-ends, or devices through API calls within the Amazon Braket SDK. The choice of device includes the various simulators and quantum computers that are available through the service. Switching from one device to another is as easy as changing a single line of code. Nonetheless, simulators are always available, whereas QPU resources might require wait time.

Some types of quantum computers are particularly well suited to solving specific sets of problems. There are many factors that determine which type of machine will meet your needs, such as qubit count, qubit fidelity (error rate), qubit connectivity, coherence time, and cost. Full specifications of the quantum computers are provided in the Amazon Braket console .

Please click here to learn more about the hardware providers of Amazon Braket. 

Analog Hamiltonian Simulation or AHS, is a different paradigm of quantum computing, distinct from universal, gate-based quantum computing. Devices capable of AHS are special purpose QPUs designed to solve a limited set of problems, represented by a Hamiltonian. AHS allows users to specify a Hamiltonian of interest, and the quantum computer tunes parameters in a way such that it emulates the continuous time-evolution of a quantum state under that Hamiltonian. As these Hamiltonians are directly implemented on the system, AHS systems do not suffer from the overhead required to formulate algorithms in terms of circuits and gates, and thus can already simulate systems with hundreds of qubits, that are prohibitive to simulate classically. Amazon Braket supports AHS via the QuEra QPUs.

Visit Rigetti’s QPU page  for system and performance information on the Rigetti QPUs, including gate fidelities and coherence times.

Visit the IonQ Best Practices  webpage for information on the IonQ QPU’s topology, gates, and best practices.

The QuEra system is a programmable QPU consisting of Rydberg atoms arranged in a two-dimensional customizable arrangement of optical tweezers. The qubits in this device consist of neutral Rubidium-87 atoms with the two states being the ground state of the atom, and one of the highly excited Rydberg state. The Rydberg-Rydberg interaction between atoms decays as the sixth power of the distance between the atoms, giving rise to an effective spin Hamiltonian with local interactions. Additionally, laser fields can tune a transverse magnetic field, to give rise to interesting spin dynamics that can produce novel ground and non-equilibrium states. Researchers accessing this device through Braket will be able to program the geometry of the two-dimensional atom arrangement, and vary the strength of the longitudinal and transverse magnetic fields in a time-dependent manner, all using the Braket SDK. This will produce an effective Hamiltonian of interest, the continuous time evolution of which can be studied on the device. For more information, refer to QuEra’s best practices here .

Quantum computing is a nascent technology, and quantum computers remain a scarce resource. Different types of quantum computers have different operational characteristics and levels of availability and so process tasks at different rates. If the QPU you select is online and not currently being used, your task will be processed immediately; otherwise it will be put in queue. As the QPU becomes available the tasks in the queue are processed in the order they were received. To notify you when your task has been completed Amazon Braket sends status change events to Amazon EventBridge. You can create a rule in EventBridge to specify an action to take, such as using the Amazon Simple Notification Service (SNS), which can send alerts to you through SMS, or other methods such as email, HTTPs, AWS Lambda or Amazon SQS.

No, not necessarily. Amazon Braket automatically compiles your code for you when you run it. However you have the option on Rigetti, OQC and IonQ devices to run your circuit as is, without compiler modifications using verbatim compilation. On Rigetti, you can additionally define only specific blocks of code to run as is, without any intervening compiler passes. To learn more, see our documentation on Verbatim Compilation .

Quantum circuit compilation transforms a quantum circuit to a compiled circuit, which undergoes qubit allocation, mapping to native gates, and optimization. However, compiler gate optimization can be problematic for researchers and quantum algorithm specialists who are developing benchmarking or error-mitigation circuits, as compiler optimizations remove or reorder gates and redundant components. With verbatim compilation, users can specify parts of circuits or entire circuits to run as is, without any compiler modifications.

Yes, you can access quantum hardware via the AWS Marketplace such as D-Wave's annealing devices .

Hybrid Jobs make the execution of hybrid quantum-classical workloads easier, faster, and more predictable. With this feature, you only have to provide your algorithm script or container, and AWS will spin up the requested resources, run the algorithm, and release the resources after completion so you only pay for what you use. The Hybrid Jobs feature also provides live insights into algorithm metrics, so you can see how your algorithm is progressing. Most importantly, jobs have priority access to the target QPU so execution can be faster, more predictable and less affected by other users’ workloads.

Braket Hybrid Jobs provide three main benefits. First, it simplifies running hybrid quantum-classical algorithms. Many quantum researchers are often new to cloud computing and do not want to set up and manage their compute environment before running their hybrid algorithm. With Hybrid Jobs, you only need to specify your preferred compute instance – or use the default. Braket Hybrid Jobs will spin up the classical resources and run the workload in pre-built container environments, return the results to Amazon S3, and finally release the compute resources so you only pay for what you use.

Second, Hybrid Jobs provide live insights into running algorithms. You can define custom algorithm metrics as part of your algorithm that will be automatically logged by Amazon CloudWatch and displayed in the Amazon Braket console. With this, you can track the progress of your algorithms.

Third, Amazon Braket Hybrid Jobs provide better performance than running hybrid algorithms from your own environment. During the entire time that your job is running, it has priority access to the selected QPU. This means tasks executed on that device as part of your job will be executed ahead of other tasks that may be queued up on the device. This results in shorter and more predictable run times for hybrid algorithms, and ultimately better results by reducing the detrimental effects of slowly changing device characteristics (‘device drift’) on algorithm performance.

You can use any of the available QPUs on Amazon Braket with Hybrid Jobs.

You can use any of the available Amazon Braket on-demand simulators (SV1, DM1, TN1), embedded simulators based on the PennyLane lightning plug-in , or a custom simulator embedded as a container to hybrid jobs. For the embedded simulators or the custom simulator, you can choose one or multiple CPU and GPU instances to run your hybrid workload.  

Embedded simulators are a set of high-performance simulators that are directly embedded in the same container as your application code to avoid latencies associated with round-trips between a fully-managed, on-demand simulator, such as SV1, and your containerized classical code. Embedded simulators support advanced features such as the adjoint method  for gradient computation which reduce the number of circuits needed to compute a gradient. Today, Amazon Braket supports embedded simulators from PennyLane, such as the lightning.gpu simulator, which is accelerated with NVIDIA cuQuantum SDK , specifically designed for running quantum circuit simulation on high-performance GPUs.

Yes, you can bring your own simulator library to Amazon Braket Hybrid Jobs by embedding the simulator and its dependencies into a container. You can then pass code to the container as an entry point and execute the code as an Amazon Braket Hybrid Job on CPU or GPU instances. Amazon Braket handles spinning up the resources for the duration of your job and you only pay for what you use.

No, by default the jobs container runs on a single ml.m5.xlarge instance type. If you are running a hybrid algorithm using an Amazon Braket on-demand simulator (SV1, TN1, DM1) or a QPU, then Amazon Braket manages the software and infrastructure for you. If you are running a hybrid algorithm using the embedded simulators from PennyLane or a custom simulator packaged as a container, then you can select one or more CPU or GPU instance types on which to run the job. Amazon Braket manages the set up of the underlying infrastructure and releases the resources once the job is complete so you only pay for what you use.

Today, the embedded state-vector simulator from PennyLane that comes pre-installed with Amazon Braket’s Hybrid Jobs container can be used for variational algorithms which can benefit from methods such as back-propagation or the adjoint method  for gradient computation. Examples of these algorithms are quantum machine learning (QML), quantum adiabatic approximate algorithm (QAOA) or variational quantum eigensolver (VQE) . With embedded simulators you also have the choice to use GPU instances, if your algorithm can benefit from GPU based acceleration, and can fit in the GPU memory. This is generally the case for variational algorithms and QML algorithms with intermediate qubit counts (< 30). Otherwise, consider using the SV1 on-demand simulator. As the adjoint method does not support non-zero shots today, consider using SV1 for any workload where the number of shots is greater than zero. Note that the embedded simulator is only supported as part of the Hybrid Jobs feature, while SV1 supports both standalone tasks and hybrid jobs. 

The PennyLane lightning.gpu simulator can be used for hybrid algorithms such as QML, QAOA or VQE, provided the problem size is small enough to fit within the GPU memory. The lightning.qubit  CPU-based simulator can be used for algorithms that are memory-intensive, and cannot fit in GPU memory, such as variational algorithms with high qubit counts (29+ qubits). Note that your costs will differ depending on whether you use a CPU or GPU instance type. Please refer to the PennyLane documentation for more details. 

For more information regarding pricing for Hybrid Jobs, refer to the Hybrid jobs tab on the Amazon Braket pricing page .

You can get started by visiting the Amazon Braket Jobs User Guide  section of Braket documentation. Amazon Braket hybrid example notebooks provide tutorials on how to get started with Jobs, and run different types of hybrid algorithms. These examples come pre-installed on Amazon Braket notebooks to help you get started quickly. You can also review hybrid algorithm examples with the PennyLane plug-in in the Amazon Braket examples repository . 

Braket Direct is an Amazon Braket program that expands the ways customers can access quantum hardware and leverage Braket to conduct cutting-edge research on today’s noisy quantum devices. Complementing existing on-demand access, customers can use Braket Direct to reserve dedicated device time for their workloads, connect with quantum computing experts to refine their workloads, and try out experimental capabilities, such as next-generation devices.

Braket Direct empowers customers looking to advance the state of the art of quantum computing – bringing them the level of device access, expert guidance, and select next-gen devices they need to accelerate their research. Braket Direct reinforces Braket's value proposition of providing a one-stop shop to access a variety of devices on a pay-as-you-go basis, without requiring upfront investments or recurring/long-term commitments.

Currently you can reserve dedicated access to IonQ’s Aria, IQM's Garnet, QuEra’s Aquila, and Rigetti’s Aspen-M-3 devices.

Yes. You can queue quantum tasks and hybrid jobs for an upcoming reservation using the Reservation ARN, or submit workloads during your reservation on the fly.

At the end of a reservation, any pending tasks/Hybrid Jobs are canceled. From there, you can choose to schedule a new reservation or re-submit for on-demand execution during public availability windows.

Once your reservation ends, you no longer have dedicated access to the device and any running Hybrid job will be canceled. We recommend using checkpoints to save and restart Hybrid jobs at your convenience.

No. Each reservation represents standalone dedicated device access. E.g., two back-to-back reservations would be considered separate and any pending tasks from the first reservation would not automatically resume in the second reservation.

Since reservations represent dedicated device access for your AWS account, the device is dedicated to you for the entire planned duration of your reservation. Therefore, you are charged for the length of the reserved time, regardless of the utilized time.

Your reservation will be canceled and you will be refunded for the remainder of your reservation time.

Reservations are available in minimum 1-hour increments and certain devices may have additional reservation length constraints (including minimum and maximum reservation durations).

Yes, you can cancel a reservation up to 48 hours in advanced of your scheduled reservation start time at no cost. Past that time, reservations can no longer be canceled and you will be charged for the entirety of the reserved time.

Yes. To reschedule, you must cancel your existing reservation at least 48 hours before the scheduled reservation start time and then create a new one.

Yes. When you create a reservation, you can sign up for an optional 30-minute reservation prep session with a Braket expert, at no additional charge.

With expert advice, you can connect with different quantum computing experts and get advice on your workload.

You can choose from Braket office hours with a Braket expert, professional services offerings from quantum hardware providers via the AWS Marketplace, or expert help from the AWS Quantum Solutions Lab.

Braket office hours are 1:1 sessions, first-come-first-serve, and take place every month. Each available office hour slot is 30 minutes and free of charge. Talking to Braket experts can help you get from ideation to execution faster: explore use case to device fit, identify options to best leverage Braket for your algorithm, and get recommendations on how to use certain Braket features like Hybrid Jobs, Braket Pulse, or Analog Hamiltonian Simulation.

No. For emergent issues or quick troubleshooting questions, we recommend reaching out to AWS Support. For non-urgent questions, you can also use the AWS re:Post forum or the Quantum Computing Stack Exchange, where you can browse previously answered

With Amazon Braket there are no upfront charges, and you only pay for the AWS resources you use. You will be billed separately for each Amazon Braket capability, including access to quantum computing hardware and on-demand simulators. You will also be billed separately for AWS services provided through Amazon Braket such as Amazon Braket managed notebooks. Please visit our pricing page to learn more about pricing. 

A: You can use tags to organize your AWS resources by logical groupings that make sense for your team or business, such as cost center, department, or project. In Amazon Braket, you can apply tags to the quantum tasks you create. After creating and applying user-defined tags, you can activate them for cost allocation tracking on the AWS Billing and Cost Management dashboard. AWS uses the tags to categorize your costs and deliver a monthly cost allocation report to you so you can track your AWS costs. Your cost allocation report displays the tag keys as additional columns with the applicable values for each row, so it's easier to track your costs if you use a consistent set of tag keys.

Yes. Scientists at universities around the world perform research on Amazon Braket using credits provided through the AWS Cloud Credit for Research program. Please submit your proposal at the link above. In the application process, if you don’t have a URL for the pricing calculator, please submit your request with a placeholder.

Yes, QPUs on Amazon Braket are hosted by our third-party quantum hardware providers. If you use Amazon Braket to access quantum computers, your circuit and associated metadata will be sent to and processed by the hardware providers outside of facilities operated by AWS. Your content is anonymized so that only content necessary to process the quantum task is sent to the quantum hardware providers. AWS account information is not transmitted to them. All data is encrypted at rest and in transit, and is only decrypted for processing. In addition, Amazon Braket hardware providers are not permitted to store or use your content for purposes other than processing your task. Once the circuit completes, the results are returned to Amazon Braket and stored in your Amazon S3 bucket. The security of Amazon Braket third-party quantum hardware providers is periodically audited to ensure standards of network security, access control, data protection, and physical security are met.

Your results will be stored in Amazon S3. In addition to providing the results of the execution, Amazon Braket also publishes event logs and performance metrics such as completion status and execution time to Amazon CloudWatch.

Amazon Braket is integrated with AWS PrivateLink, so you can access Amazon Braket from within your Amazon Virtual Private Cloud (Amazon VPC) without requiring the traffic to traverse across the Internet. This reduces exposure to security threats from Internet-based attacks and the risk of sensitive data leakage.

The Amazon Quantum Solutions Lab is a collaborative research and professional services program staffed with quantum computing experts who can assist you to more effectively explore quantum computing and work to overcome the challenges that arise with this nascent technology. Please visit the Quantum Solutions Lab webpage to get started. 

You can request information about engagements with the QSL and our partners by submitting this form , and by working through your AWS account manager.

The cost of QSL engagements vary depending on the length of the engagement and nature of your needs. Please reach out to your account manager for more details. 

Quantum Solutions Lab engagements typically last from 6 to 12 months.

The entire process can be done remotely, if needed, which is likely during the current pandemic. However, typically we meet in person to kick off the engagements and determine a working cadence. After that, we will visit your site as needed and have regular checkpoints using video conferencing, while collaborating remotely on a regular basis.