Xanadu has released new versions of PennyLane and Catalyst, featuring several improvements and new features. The update includes better integration with Qiskit 1.0, making it easier to import workflows from Qiskit into PennyLane. The release also introduces CUDA Quantum integration, allowing users to compile their PennyLane workflow using CUDA Quantum. Other features include native mid-circuit measurements on default.qubit, a new Clifford device, and enhancements to Catalyst, such as vectorization support with vmap and improved mid-circuit measurement capabilities. The team behind the update includes Isaac De Vlugt from Xanadu and Thomas Bromley and Josh Izaac from the PennyLane Team.
PennyLane v0.35 and Catalyst v0.5: New Features and Improvements
The latest versions of PennyLane and Catalyst have been released, introducing a range of new features and improvements. These updates aim to enhance the user experience and expand the capabilities of these quantum computing software tools.
Integration with Qiskit 1.0 and CUDA Quantum
One of the key updates in PennyLane v0.35 is the improved integration with Qiskit 1.0. This enhancement simplifies the process of importing workflows from Qiskit into PennyLane. The function qml.from_qiskit converts a Qiskit QuantumCircuit into a PennyLane quantum function. This function has been improved to make importing from Qiskit easier and is compatible with both Qiskit 1.0 and earlier versions.
In addition to full circuits, it is now also possible to convert operators from Qiskit to PennyLane with a new function called qml.from_qiskit_op. This function allows a Qiskit SparsePauliOp to be converted to a PennyLane operator.
Another significant update is the integration of CUDA Quantum with PennyLane. When using PennyLane’s @qml.qjit decorator for just-in-time compilation, users can now specify @qml.qjit(compiler="cuda_quantum") to utilize CUDA Quantum to compile their PennyLane workflow and execute it on CUDA Quantum supported backends.
Mid-Circuit Measurements and Clifford Device
PennyLane v0.35 also introduces improvements to mid-circuit measurements (MCMs). MCMs can now be made more scalable and efficient in finite-shots mode with default.qubit by simulating them in a way similar to what happens on quantum hardware. This approach works well when there are a lot of MCMs and the number of shots is not too high.
A new default.clifford device has been added to PennyLane, enabling efficient simulation of large-scale Clifford circuits. This device uses Stim as a backend and can provide the usual range of PennyLane measurements as well as the state represented in the Tableau form.
Catalyst Features: Vectorization, Post-Selection, and Qubit Reset
Catalyst v0.5 introduces vectorization support with vmap, making it easier to modify functions to map over inputs with additional batch dimensions. This feature is designed to minimize the usage of Python for loops and push as much of the computation through to the array manipulation library.
Mid-circuit measurements in Catalyst now support post-selection and qubit reset when used with the Lightning simulators. To specify post-selection, users can pass the postselect argument to the catalyst.measure function. To reset a wire after mid-circuit measurement, users can specify reset=True.
Other Improvements
In addition to the new features, the release contains a wide array of improvements and optimizations. These include improved operator arithmetic functionality, support for CUDA version 12 in Lightning GPU, faster computations with Vector-Jacobian products (VJPs), and more flexibility for gradient functions used within a @qjit compiled function.
The @qml.qjit decorator has also been improved, reducing the need to recompile the function if previously called with known argument types and static values. Furthermore, capturing quantum circuits for just-in-time compilation with @qml.qjit with many gates is now quadratically faster.
External Link: Click Here For More
