Pennylane-forest

New features since last release

Re-introduction of the Rigetti Quantum Proccessing Unit (QPU) device

This release uses the latest version of [pyQuil](https://github.com/rigetti/pyquil) to connect to [Rigetti Quantum Cloud Services (QCS)](https://docs.rigetti.com/qcs/) and enables the use of the latest [Rigetti QPUs](https://qcs.rigetti.com/qpus) as a PennyLane device. [#107](https://github.com/PennyLaneAI/pennylane-forest/pull/107)

Breaking changes

* The package has been renamed to `pennylane-rigetti`. The top level import is now `pennylane_rigetti`. In addition, the prefix for device short names have been changed to `rigetti`. For example, `rigetti.qpu`. [110](https://github.com/PennyLaneAI/pennylane-forest/pull/110)

* A new version of the QCS CLI is required if you want to use your QCS account to run your workloads on a live Rigetti QPU. See [Using the QCS CLI](https://docs.rigetti.com/qcs/guides/using-the-qcs-cli) for details. [#107](https://github.com/PennyLaneAI/pennylane-forest/pull/107)

* The `forest_url` parameter has been removed, as it is now managed by the QCS CLI. [107](https://github.com/PennyLaneAI/pennylane-forest/pull/107)

* The `compiler_url` and `qvm_url` device parameters have been removed. The default URLs can be overridden using the `QCS_SETTINGS_APPLICATIONS_PYQUIL_QUILC_URL` and `QCS_SETTINGS_APPLICATIONS_PYQUIL_QVM_URL` environment variables, respectively. [107](https://github.com/PennyLaneAI/pennylane-forest/pull/107)

* The `timeout` parameter for all devices has been renamed to `compiler_timeout`. [107](https://github.com/PennyLaneAI/pennylane-forest/pull/107)

* There is now a default execution timeout of 10 seconds. This can be configured for a device by using the new `execution_timeout` parameter. [107](https://github.com/PennyLaneAI/pennylane-forest/pull/107)

* The `S`, `T`, `CSWAP`, `ISWAP`, and `CCNOT` operations have been removed. Import them directly from `pennylane` instead.[107](https://github.com/PennyLaneAI/pennylane-forest/pull/107)

Improvements

* Improves the computation of the expectation value when using `QPUDevice` by skipping the `Device.generate_samples` method. [108](https://github.com/PennyLaneAI/pennylane-forest/pull/108)

Bug fixes

* The QPU device now correctly sets the number of shots when parametric compilation is disabled. [107](https://github.com/PennyLaneAI/pennylane-forest/pull/107)

Contributors

This release contains contributions from (in alphabetical order):

Albert Mitjans Coma, Antal Szava, Marquess Valdez.

Bug fixes

* Defines the missing `state` method and `returns_state` entry of the `capabilities` dictionary for `forest.wavefunction` and `forest.numpy_wavefunction`. [(36)](https://github.com/PennyLaneAI/pennylane-qulacs/pull/36)

Contributors

This release contains contributions from (in alphabetical order):

Antal Száva.

Bug fixes

* Fix a bug where array parameters where not accepted when building circuits on pyQuil side. [(90)](https://github.com/PennyLaneAI/pennylane-forest/pull/90)

Improvements

* Added support for Python 3.10. [(96)](https://github.com/PennyLaneAI/pennylane-forest/pull/96)

Contributors

This release contains contributions from (in alphabetical order):

Romain Moyard.

Breaking changes

* Deprecated Python 3.6. [(85)](https://github.com/PennyLaneAI/pennylane-forest/pull/85)

Improvements

* Removed a validation check for ``QubitUnitary`` that is now in PennyLane core. [(74)](https://github.com/PennyLaneAI/pennylane-forest/pull/74)

Bug fixes

* Pins the PyQuil version to use as `pyquil>=2.16,<2.28.3` due to API deprecations in PyQuil version 3.0. [(73)](https://github.com/PennyLaneAI/pennylane-forest/pull/77)

Contributors

This release contains contributions from (in alphabetical order):

Theodor Isacsson, Romain Moyard, Antal Száva.

Improvements

* Adds support for Python 3.9. [(74)](https://github.com/PennyLaneAI/pennylane-forest/pull/74)

Bug fixes

* Fixed a bug caused by the `expand_state` method always assuming that inactive wires are the least significant bits. [(73)](https://github.com/PennyLaneAI/pennylane-forest/pull/73)

Contributors

This release contains contributions from (in alphabetical order):

Antal Száva.

Breaking changes

* For compatibility with PennyLane v0.15, the `analytic` keyword argument has been removed from all devices. Analytic expectation values can
still be computed by setting `shots=None`. [(71)](https://github.com/PennyLaneAI/pennylane-forest/pull/71)

* For compatibility with PennyLane v0.15, parametric compilation now depends on the `requires_grad` attribute of parameters instead of the deprecated `Variable` class. [(71)](https://github.com/PennyLaneAI/pennylane-forest/pull/71)

* The circuit hashes used for parametric compilation are computed in `QVMDevice` instead of in `QubitDevice` defined in Pennylane. [(71)](https://github.com/PennyLaneAI/pennylane-forest/pull/71)

Contributors

This release contains contributions from (in alphabetical order):

Antal Száva.