New features since last release
* Add MCMC sampler.
[(384)](https://github.com/PennyLaneAI/pennylane-lightning/pull/384)
* Serialize PennyLane's arithmetic operators when they are used as observables
that are expressed in the Pauli basis.
[(424)](https://github.com/PennyLaneAI/pennylane-lightning/pull/424)
Breaking changes
* Lightning now works with the new return types specification that is now default in PennyLane.
See [the PennyLane `qml.enable_return`](https://docs.pennylane.ai/en/stable/code/api/pennylane.enable_return.html?highlight=enable_return) documentation for more information on this change.
[(427)](https://github.com/PennyLaneAI/pennylane-lightning/pull/427)
Instead of creating potentially ragged numpy array, devices and `QNode`'s now return an object of the same type as that
returned by the quantum function.
>>> dev = qml.device('lightning.qubit', wires=1)
>>> qml.qnode(dev, diff_method="adjoint")
... def circuit(x):
... qml.RX(x, wires=0)
... return qml.expval(qml.PauliY(0)), qml.expval(qml.PauliZ(0))
>>> x = qml.numpy.array(0.5)
>>> circuit(qml.numpy.array(0.5))
(array(-0.47942554), array(0.87758256))
Interfaces like Jax or Torch handle tuple outputs without issues:
>>> jax.jacobian(circuit)(jax.numpy.array(0.5))
(Array(-0.87758255, dtype=float32, weak_type=True),
Array(-0.47942555, dtype=float32, weak_type=True))
Autograd cannot differentiate an output tuple, so results must be converted to an array before
use with `qml.jacobian`:
>>> qml.jacobian(lambda y: qml.numpy.array(circuit(y)))(x)