PennyLane from Xanadu Releases New Quantum Computing Features and Improvements

PennyLane, from Xanau leading quantum computing software company, has released new features and improvements to its platform. The updates include support for just-in-time compilation workflows with the Qrack GPU simulator, allowing for faster execution speeds and dynamic quantum control flow. Additionally, PennyLane now supports arbitrary state-based measurements with qml.Snapshot and has updated its implementation of qml.assert_equal for various operator instances.

The company also announced improvements to its Catalyst compiler, including support for autodifferentiation and custom VJP rules. Furthermore, PennyLane’s QChem Hamiltonians can now be converted to and from OpenFermion with qml.to_openfermion and qml.from_openfermion.

These new features were developed in conjunction with UnitaryHACK 2024, a hackathon event. Key companies involved in this work include PennyLane, Harvard, and Qrack.

New Features

1. Arbitrary state-based measurements: The default.clifford module now supports arbitrary state-based measurements with qml.Snapshot.
2. Updated implementation for qml.assert_equal: The implementation has been updated for various operator instances, including Operator, Controlled, Adjoint, Pow, Exp, SProd, ControlledSequence, Prod, Sum, Tensor, and Hamiltonian.
3. Enhanced qml.from_qasm functionality: It now supports converting mid-circuit measurements from OpenQASM 2 code and takes an optional argument to specify a list of measurements to be performed at the end of the circuit.
4. New operator: qml.QutritChannel: This allows for specifying noise on the default.qutrit.mixed device using a collection of (3×3) Kraus matrices.

Improvements

1. Efficient gradient computation with catalyst.value_and_grad: Returns both the result of a function and its gradient with a single forward and backward pass, reducing overall quantum executions.
2. Dynamic one-shot method for simulating circuits with mid-circuit measurements: Catalyst now supports this method, which can be advantageous for circuits with many mid-circuit measurements executed for few shots.
3. AutoGraph enhancements:
   • Single-index in-place array assignments are now captured and automatically converted into JAX-compatible form.
   • The catalyst.disable_autograph decorator allows disabling AutoGraph for specific external functions.
   • Arrays with dynamic shapes can be used with control-flow primitives like for_loop, while_loop, and cond.
4. QChem Hamiltonians conversion: QChem Hamiltonians can now be converted to and from OpenFermion using qml.to_openfermion and qml.from_openfermion.
5. QROM algorithm availability: The QROM algorithm is now available in PennyLane with qml.QROM, allowing for entering classical data in the form of bitstrings.

Deprecations and Breaking Changes

1. Simplify argument deprecation: The simplify argument in qml.Hamiltonian and qml.ops.LinearCombination has been deprecated; use qml.simplify instead.
2. Default behavior change for qml.from_qasm(): The default behavior of removing measurements in the QASM code has been deprecated; use measurements=[] to keep this behavior or measurements=None to keep the measurements from the QASM code.
3. Removal of qml.load and qml.from_qasm_file: These functions have been removed in favor of more specific functions.

These updates demonstrate the continued growth and development of PennyLane and Catalyst, making them even more powerful tools for quantum computing and machine learning research.